R&D Resources
SMI Research Spotlight Series: Practical Close Encounter Navigational Engine
The COLREGs (International Regulations for Preventing Collisions at Sea) rules are not specifically structured to address situation involving complex navigational scenarios involving multiple vessels, To address this critical safety issue in the shipping industry, a multi-disciplinary SMI-funded Centres of Excellence project team comprising researchers from C4NGP (NUS), CEMS (SP) and CEAOPS (TCOMS) came together to collaborate and work with PSA Marine to propose practical navigational guidelines in multi-ship encounters. Utilising a state-of-the-art deep reinforcement learning (DRL) model, the team developed multi-ship collision avoidance recommendations and tested them on an agent-based simulator that can simulate a diverse range of vessel types found in Singapore waters. The results demonstrated that the DRL model can significantly reduce the likelihood of collisions and casualties caused by human error in complex multi-ship conflict situations.