This week, we dedicated our lab lesson on Monday to building the structure of the boat using a plastic bowl and foam board. As mentioned last week, while we were originally planning on starting with a foam raft and later implementing a foam board boat, Professor Gloria suggested the bowl as a way to simplify our design as we know it is buoyant and waterproof. For additional stability, and to help protect our electronics, we added foam board to the sides using a waterproof sealant. We have not attached our motors to the board yet, as they are not waterproof (boo false advertising), however we have ordered new brushed motors. As a group, we feel that we are in a good place for the upcoming interim demo. Our next steps will include waterproofing our electronics and attaching the boat to the underwater unit.

Emma spent time on the motor control and video streaming, and was able to get both working successfully! Last week, there were several bugs in the implementation, which resulted in the socket server of the Raspberry Pi failing to run the livestream and control the motors at the same time. This week, she separated the motor control from the video feed, using an Arduino Nano and Arduino Cloud to remotely control the motors over Wi-Fi at a distance of 10 feet, which is a requirement our project. She was also able to improve the framerate and color clarity for the video streaming on the Raspberry Pi and pick out new motors for our boat.

Abie spent time on the ML algorithm and was able to implement code that used a KNN model from an external library (scikit-learn) to classify groups of pixels as one of two coral types, pink or blue, and then classify each coral color type as one of four categories: bleached, partially bleached, pale, and healthy. Her next steps include adding more training data for the model and increasing the ability of the model to exclude the background from the sections being classified.

Maddie spent time engineering a waterproof sleeving for our underwater cables and redesigning the spool that we will use to control the z-direction of our project. Our stepper motor and driver came in this week, so she was also able to finally set-up the stepper motor with position tracking code as well as continue working on the image recombination code.

This week, our team has set a goal of being able to combine the subsystems so we have end to end operation, even if there are some bugs, by the end of next week (Friday the 21st). In order to realistically achieve this goal, we’ve set pre-MVPs for each component, and have been making progress towards achieving each of these separate goals. We are all currently debugging specific parts of the project, and will be integrating our subsystems during the week. 

After meeting with Professor Gloria, she suggested that we should try and use a plastic bowl for our above water boat to start with, instead of constructing our own initially. Plastic is inherently waterproof, so we would only need to cut holes for the motors and fill the edges with waterproof paste. This will protect our electronics in the early stages of prototyping and demonstrations. We can improve this by purchasing a premade boat or constructing our own, but for our interim demo, we will probably use a simpler plastic design.

Maddie primarily spent time this week working on taking video and extracting overlapping-enough images to be used in creating the map. This functionality is currently entirely separate from the actual interface, but is functioning and accurate enough for initial implementation. Maddie has also been re-working the Z-direction movement system; the new design is simpler and will likely work better, however, has the potential to negatively impact balance of the boat. She will be prototyping the simplified spool design (more details in personal status report) for the purposes of the pre-MVP, and is confident that this will be enough testing to be able to finalize the design. Maddie has already pre-emptively updated (as a separate code version) the stepper motor code to accommodate the redesign we’re considering/testing this upcoming week.

Emma spent time this week working on the interface between the computer and Raspberry Pi. This week we were focused on getting the back end of the interface working before making GUI improvements. Currently, the back end connection is being made from a socket server. The pi is sending image frames in a loop over the socket, and the computer is sending GPIO requests to control the motors for lateral movement. There are bugs in the implementation right now that are being investigated, mainly that since the frames are being sent and requested so frequently, motor control signals are not being processed. A lot of time has been spent improving the system, but this is still being worked on. 

Abie spent time this week working with the pillow library in python for image processing to group and identify the RGB values of different pixel clusters for classification. Before next Friday, this needs to be implemented with a basic KNN model. Additionally, Abie plans to work with Emma to figure out how to export the camera pictures in a way that is convenient for the code to use. One hurdle with the current code is efficiently processing a large set of images, so this process will need to be updated once it is integrated with the other project systems.