Autonomous Surface Vessel Design Team
I am currently the director of autonomy for Aquatonomous, a design team building an autonomous surface vessel for ecological research and for the Roboboat competition.
My team and I are responsible for the boat's computer vision, mapping, and path planning.
My contributions are mainly to the design of the autonomy pipeline, administering the team's ROS2 environment including MAVROS and launch files, and using the boat's microphone to recognize sound signals.
At a high level, the boat's autonomy pipeline has computer vision, mapping, and path planning stages.
The computer vision stage uses a YOLO model to classify buoys and other boats.
The mapping stage uses DBSCAN clustering on the LIDAR pointcloud to detect and localize obstacles.
Data from computer vision is then used to classify obstacles.
A behavior tree is used to determine the next waypoint the boat must reach to complete its tasks.
A potential fields algorithm is used to navigate to that location.
In the previous year I was a general member of the autonomy team and worked on path planning.