The Research Assistant Module for Subsea Exploration and Analysis (RAMSea) is a scalable, low-cost underwater environmental monitoring system that integrates artificial intelligence (AI), data processing, and water quality sensors. Housed within a small watertight enclosure and designed to be either operated by a diver, or controlled remotely, the system integrates AI for real-time fish recognition, implemented on a Raspberry Pi using the YOLOv8 deep learning algorithms. External to the enclosure are a suite of environmental sensors measuring depth, temperature, salinity, and dissolved oxygen. By correlating fish species observations with environmental parameters directly in the field, RAMSea provides a simple but comprehensive approach to marine ecosystem analysis. The system will offer a significant improvement over traditional manual survey methods by providing richer, real-time data to inform environmental management, aquaculture practices, and biodiversity monitoring.
Initiated through a student internship program in May 2025, RAMSea has been designed to provide an innovative monitoring solution for the current Harmful Algal Bloom (HAB) impacting over 500 km of South Australia’s coastline. Over the period of just a few months, this natural disaster has seen large numbers of dead marine animals washed up across South Australian coastlines, and has resulted in unprecedented impact on commercial and recreational fishing industries within the State. When fully developed, the RAMSea system could provide an effective solution to monitor the impact and evolution of this type of event. Not limited to the monitoring of HABs, RAMSea could also be applied to a multitude of other applications such as supporting reef conservation programs in the Great Barrier Reef, enhancing aquaculture health monitoring, and contributing to long-term climate impact studies.
While initial deployment of RAMSea will be via a trained scuba diver, the project team is concurrently developing a remotely operated vehicle (ROV) that will carry the sensor module as a payload, thus alleviating the need for a trained diver. A future enhancement to the system will include an autonomous surface vehicle (ASV) capable of autonomously controlling the ROV and sensor module from the surface while also communicating collected data to the cloud. This flexible configuration will deliver a modular and adaptable platform capable of operating in diver-operated, remotely operated, or fully autonomous modes.
The first prototype of RAMSea has been designed and constructed. Several tests have been conducted including qualifying the waterproof enclosure in a test tank, and sea testing from a jetty to validate the environmental sensors, the camera and the control code. Due to the low visibility in the water during the jetty test program, the fish recognition system could not be tested in the field. Instead, the real-time fish recognition system has undergone initial testing in a simulated environment involving various videos of fish in their natural environment. Field testing of the fish recognition system will take place in the upcoming test program, currently scheduled for March 2026. The results obtained from these trials have been promising and have already contributed to iterative improvements in both hardware and software and have validated the usability of the system.
