Cornell AutoBoat

We are a group of passionate and hard-working engineers who are committed to innovating in the field of maritime technology. Our goal is to design, manufacture, test, and compete with an autonomous boat capable of decision making and complex path planning using computer vision. We compete in the annual Roboboat competition, run by Robonation.

Check out our team video!

Competition Information

The Design


In the research phase, we familiarized ourselves with ship engineering and fluids as well as concurrent literature on the subject.


In the design phase, we optimized performance by doing trade studies and cost-benefit analyses to drive design decisions.


In the manufacturing phase, We have developed a hybrid fiberglass-foam hull manufacturing method for this year’s competition.


We conduct many tests on our boat’s different subsystems, including testing software in a simulator to running practice courses at the pool to test the integration of each component.


Before beginning the design process, we first needed to research past, present, and future efforts in maritime engineering and autonomous robotics. We focused on answering our questions with underlying first principles and fundamental concepts, which broke down the complex problem into smaller, simpler tasks. In the research phase, we performed a thorough literature review of books and papers on ship engineering, fluids concepts, electrical safety, autonomous navigation algorithms, computer vision models, and more. Here we also discussed potential design methods tradeoffs, assessing the viability of various techniques. The research phase not only gave us the fundamental tools to mechanically and electrically construct the boat, but it also prepared us to incorporate advanced autonomy into our design. We maintain documentation of our findings and valuable insights from this research phase, which is particularly useful in onboarding new members. We plan on extending our research for years to come.


This year, our primary design objective is to maximize stability, in order to minimize the disturbances that the computer vision camera experiences while completing tasks. After considering the pros and cons of different designs, we settled on a trimaran design with one main hull in the center, and two smaller stabilizing hulls on either side. The design process includes a lot of CAD work and complex analysis in programs such as Ansys fluent and Fusion 360 in order to simulate how different designs would behave in the water, and some byhand mathematical calculations to estimate the waterline.

On the software side, the design phase consists of brainstorming creative task completion strategies, experimenting with different object recognition model configurations, and determining the structure of our ROS framework. In this phase interdisciplinary communication is crucial as many mechanical, electrical, and software considerations affect one another.


Our hulls for this year’s boat are made out of fiberglass with a closed-cell foam core on the inside. In order to create the specific shapes for the hulls, 2-D outlines for different side profiles were cut out from cardboard, and the foam cores were meticulously sanded down to match all the precise curvature in the design. Fiberglass was then cut out and laid over the foam, saturated in laminating epoxy resin. The wet, layup cures overnight, after which all of the sharp edges are sanded down. A body filler layer is added to provide some more stiffness to the fiberglass, and a finishing gelcoat is also applied. Now the hulls are ready to be painted and integrated with all of the electrical and robotic components. The main hull’s foam core is taken out, and wooden ribs are inserted, which provide lateral support and a structure in which the electrical components reside. A wooden board is cut to size and attached on top of the ribs, to serve as the bridge deck on top of which the robotic components and cameras are placed. Read more about the manufacturing for robotics by checking out the mechanical subteam page!


The testing phase was an integral part of our development plan, understanding that verifying and correcting the behavior of our designs is critical to our success. Our testing plan consists of simulation, live-camera, and in-water components. The simulation testing is purely software. For the task completion algorithms, simulation testing consisted of an expansive test suite for each challenge, verifying the correct course of action is planned. For the object detection system, we analyzed the accuracy of our trained models. We also manually and visually verified our models with live camera tests and ensured that the detection and position data provided by the ZED camera was accurate. The in-water testing is a full-system software and hardware test, and was the final phase of our testing. In this stage we test the full autonomous capabilities of the boat and evaluate how the mechanical, electrical, and software components can seamlessly integrate into a single product.


Giving Day