Perception and Motion Planning for Autonomous Surface Vehicles in Aquaculture

Abstract

The "Oystermaran" USV was designed and developed by students at MIT SeaGrant to resolve the oyster basket flipping bottleneck that slows down oyster farming at Ward Aquafarms. The state of the USV requires remote operation within close distance of the vessel. In this thesis, we present an automated solution that will enable the Oystermaran to autonomously depart from its parked location, navigate to its destination, execute the flipping tasks, and return to a designated location with little to no human intervention. The details explored in this project and discussed in the thesis focus on the perception and motion planning aspects of the proposed autonomous system. Our results show a capable basket detection algorithm based on our collected dataset. The system’s path planning approach is also proven sufficient in simulation. Additional data collection with further testing may be required to fully realize the system on board the Oystermaran.