Our AUV's Vision and Navigation system

Our capstone team has been making exciting progress in developing our autonomous underwater vehicle (AUV) for the RoboSub 2025 competition. Over the past few weeks, we've been setting up the critical software and hardware components that will allow our AUV to navigate, recognize objects, and complete competition challenges.

So far, we have successfully:

• Set up Isaac ROS Common, which provides essential tools for running AI-powered robotics applications.

• Configured DeepStream with YOLO, calibrated with our camera for real-time object detection.

• Integrated Pixhawk, which will handle the AUV’s navigation and control.

• Set up GXF (Graph eXecution Framework) to manage compute pipelines efficiently.

• Identified a balloon as a frisbee using YOLO, a key step in fine-tuning our object recognition model.

While we’ve made great strides in setting up the software, there’s still a lot to refine. Our next steps will focus on:

• Fine-tuning YOLO to correctly classify underwater objects.

• Enhancing image processing to adapt to underwater conditions.

• Testing autonomous movement using the Pixhawk controller.