Thomas Scherlis - tscherli@andrew.cmu.edu

TAUV is designing a system to autonomously navigate a series of competition tasks including navigating gates, manipulating objects, and responding to acoustic signals. The aquatic platform will feature a custom real-time embedded electronics system, an NVIDIA Jetson computer, and a hot-swappable smart battery. Using several cameras and computer-vision based object recognition, we can accurately detect competition objects and plan a route through the pool without relying on expensive doppler systems. We hope to place competitively in our first year at the RoboSub competition.


We use ROS to manage our high-level navigation control and simulate using Gazebo. To detect competition objects, we use the YOLOv3 machine learning based computer vision library. The software will create a map of the competition field and navigate accordingly.

We use Python and ROS.


TAUV features an NVIDIA Jetson as well as a system of 3 custom onboard controllers: a high-speed control board designed for digital signal processing, a supplemental power-management board to power actuators and the electronics system, and a smart, hot-swappable battery pack with state-of-charge estimation and protection features.

We design custom PCBs for all of our electronics, and program them using FreeRTOS and C.


TAUV’s mechanical system is based on an acrylic pressure-hull for the electronics platform, with 8 powerful thrusters for 6 degrees of freedom control. We have a light-weight aluminum superstructure with laser-cut acrylic panels. Batteries are hot-swappable and can be charged without opening the seal.