Website under construction! Stay tuned!

Under Construction

This page is still under development. ARVP's website development team is working hard to provide the information you're looking for. We pride ourselves on making a beautiful and functional website and appreciate your patience.

In the meantime, check us out on Instagram! @uofa.arvp

ARCTOS Overview

A Beautiful Fusion of Electrical, Mechanical and Software Engineering. With Arctos, ARVP prioritized functionality and modularity in a minimalist design. Artocs employs a rectangular hull machined from bonded aluminum plates with an acrylic front cap, allowing camera vision, and a top lid to access batteries. Both access points are held in place by aluminum straps to maintain o-ring seals. External wiring is routed via penetrators on the hull’s rear wall. Arctos’ frame is made from stock L-brackets and welded aluminum tubing, creating excess mounting space to test new subsystem designs and configurations.

In addition to a Jetson Orin and other stock components mentioned in Appendix A, Arctos’ electrical trays harbor the battery monitoring and carrier board, which draw power from five Lipo batteries and convert it into 5V and 12V rails. The communications hub, internal environment, and actuator boards transfer, display, and action data across the platform, while the sonar boards handle pinger detection.

The new software stack is entirely executed in Docker containers, creating consistent and reproducible environments across different machines. Additionally, core dependencies have been updated such as Ubuntu 18.04 to Ubuntu 20.04 and ROS 1 melodic to ROS 2 Humble.

Software Overview

Computer vision plays a huge role in sensing objects underwater autonomously; every bit of performance has knock-on effects for the capabilities and robustness of our entire autonomous system. Previously, we used transfer learning fine-tuning on YoloV3 models to create our own custom object detecting models. Which, were then fine tuned with labeled data sets for bounding box detection.

This year, we transferred to a faster and more accurate base model version, YoloV7, to derive from. To further optimize performance we compiled these models with Nvidia's TensorRT compiler. This allowed an increase from 15 fps to 60 fps of real time image processing run on a Nvidia Jetson AGX Orin.

While our previous dynamics control system allowed full control over position and velocity in 3D, it was slow and difficult to improve incrementally. This year, we bettered our dynamics control by removing all assumptions and starting ground up; emphasizing simplicity in design and only being as complex as necessary. We decided to make a Cascading PID controller. Through physical demonstration tests, we found our new simple control system much more responsive and robust dynamic feedback system; a significant improvement over previous years.

Mechanical Overview

Hull

Droppers

Claw

Torpedoes

Electrical Overview

ROBOT HISTORY

Arctos
Top result: 2nd in hull design
2020-2023
Auri
Top Result: 4th overall in 2019 and 2022
2017
AquaURSA
Top Result: 8th overall
2013
SubmURSA
Top Result: 8th overall
2011
Bearacuda
Top Result: 11th overall
2008
linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram