Alfie, our AUV, was designed, built, and programmed according to three central design principles: simplicity, scalability, and stability.
Simplicity acknowledges that systems which minimize complexity are the most likely to succeed.
Scalability urges engineers to consider the future and design a platform which can grow with the team.
Stability encourages designs that favor longevity, so that existing systems will not need to be replaced.
Alfie consists of a simple welded box-shaped aluminum enclosure mounted to a base plate. The various actuators and sensors which Alfie uses to interact with its environment are mounted on the base plate. Eight thrusters allow Alfie to move in six degrees of freedom. The core mechanical design makes Alfie very physically stable. It sits flat in the water, and naturally tends to move in a straight line, unless instructed otherwise.
In RoboSub, one of the hardest tasks is to launch a torpedo through a small hole in a target. Our torpedo subsystem launches a 3D printed projectile using a spring. This simple mechanism allows the torpedo to travel straight in the water.
The RoboSub course also includes several objects which must be picked up and dropped in a new location. Alfie's has a gripper for this task, which uses a servo to actuate fingers to pick up the various game-pieces. This gripper is mounted to an arm which allows more flexibility in the positioning of Alfie.
MuddSub's current mechanical work on Alfie is supported by 3 primary subteams: force fixture, gripper, and torpedo.
The electrical team aims to build the infrastructure and sensor systems to support Alfie's software stack. Using custom-designed PCBs, Alfie is able to communicate with all the sensors, deliver reliable power to each subsystem, and use control actuators to interact with the environment.
Members of the electrical team work to build a reliable, safe, and robust robot which makes full use of the mechanical systems and integrates tightly with software. Members use industry-grade tools such as Altium and LT Spice, providing valuable real-world experience in electrical engineering.
The MuddSub software team aims to make Alfie fully autonomous. The team approaches this from several directions: perception, planning, navigation and control, and infrastructure. In the perception division, Computer Vision team extracts landmarks’ poses from images, while SLAM creates a map and localizes Alfie. Motion Planning and Control charts paths and computes thruster forces, ensuring Alfie’s safe travel. Finally, Planning and Infrastructure acts as Alfie’s information hub, providing architectural and mission planning and simulations.
Team members gain experiences in ROS, PyTorch, Gazebo, and more. They apply important algorithms such as YOLO and FastSlam, and learn about advanced fields such as deep learning and reinforcement learning.