Technical Report: here!
Website: here!
Competition Website: here!
TartanAUV is Carnegie Mellon University’s underwater robotics team that competes in the international Robosub competition every year. We were founded in 2018, and this is our second year developing our autonomous vehicle.
I am the Software Lead for TartanAUV’s 2019-2020 season. My role encompasses organization, design, and execution of the software pipelines. My focus for this season was to implement a functional perception pipeline. This includes detectors, data association, tracking, and a backend SLAM system to localize both the robot and the landmarks. I draw upon my classwork at CMU (SLAM, ML, Computer Vision) and was able to put many of the advanced concepts covered in those classes to use in the real world.
Below is this season’s competition video. Since the competition was cancelled due to COVID-19, we submitted a paper and video and will continue developing the sub for next year’s competition. I hope to continue contributing to the team and advance Carnegie Mellon in the Robosub competition!
My goals for the next season/near future are as follows:
Implement an efficient loop closure system using features from the floor/landmarks
Implement a bearing-only SLAM system that is robust to the inaccuracies of stereo vision underwater
Work on semantic navigation (move to the right of the buoy) instead of having to enumerate and hard-code navigation patterns
Implement a dense mapping system (not even necessary for this application but still cool!)