| Summary: | With the global population on the rise, there is an increased demand for seafood, underscoring the crucial role of aquaculture- the practice of farming aquatic organisms [1]. In the realm of aquaculture, oyster farming is relatively low maintenance, except for the challenge of manually flipping heavy oyster-laden bags. To address this issue, MIT Sea Grant introduced the Oystermaran, an autonomous catamaran specifically designed for this task. This thesis presents contributions to the electronics, controls, and perception systems of the Oystermaran project. In particular, it presents an oyster basket detection and tracking method using the object detector You Only Look Once (YOLO) [2]. In addition, the electronics system has been updated and new manual controllers were created to enable the use of a new f lipping mechanism developed this year. This system is evaluated on data from field testing at Ward Aquafarms, a Cape Cod-based oyster farming business. The results show that oyster baskets can be robustly detected in new environments, despite environmental factors. This marks a significant step towards real-time viability for autonomous oyster farming.
|
|---|