Competition strategy and team progress
Rising from the ashes of defeat and disorder following the pandemic, the phoenix soars to competition this year with renewed pride, a greater sense of togetherness, and a singular focus: to compete hard while having fun.
After a hard-fought competition the previous year, the team chose to focus on four principles for this year: decoupling, reliability, maintainability, and performance. Not only was the AI pipeline decoupled and tested separately, but nearly all major components of the boat could be separated, making it extremely modular and maintainable. It was the first year that the team tested in the Marine Hydrodynamics Laboratory, and did so for over 100 hours. In addition, it was the first year that the boat was made out of carbon fiber from scratch - from the painstaking mold making, preparation, and vacuum resin infusion - in order to produce a boat that weighed less than 60 pounds while still delivering impeccable results.
The Phoenix weighs 55 pounds, has a length of 56”, a beam (width) of 30”, and has an overall height of 26”. It can generate thrusts of more than 25 pounds. A hull and superstructure made of carbon fiber allows for weight reduction which has the benefits of an increased thrust to weight ratio and ease of transportability.
An AMD Ryzen 5 5600X on a ASUS ROG Strix X570-I motherboard with 32 GB RAM provides the main logic of the boat. A Ubiquiti Rocket point-to-point transceiver allows for a reliable, fast data connection to land control. An Arduino Mega is used for computer control of the two Blue Robotics T500 thrusters. A Velodyne Puck LiDAR paired with a Logitech webcam is used for the boat's sensing.
A Docker-based software environment configured with ROS 1 Noetic ensures that the exact same software is present when testing on a computer and on the boat—allowing team members to use their choice of Linux, Windows, or macOS on their personal computers without differences. The team uses Git as our version control system. Most of the team's code is written in C++, with several Python scripts for small tasks. Computer Vision is done with the YOLOv4 object detection system, and an upgrade to YOLOv8 is planned for the 2023-2024 year. The team plans to migrate to ROS 2 as well this year.