Dong, X-Y and Cai, W, 2024. Collision avoidance decision making for ships based on DRL. In: Phillips, M.R.; Al-Naemi, S., and Duarte, C.M. (eds.), Coastlines under Global Change: Proceedings from the International Coastal Symposium (ICS) 2024 (Doha, Qatar). Journal of Coastal Research, Special Issue No. 113, pp. 829-833. Charlotte (North Carolina), ISSN 0749-0208.

In congested maritime areas and complex navigational environments, ensuring safety through effective collision avoidance is paramount. Traditional methods, largely reliant on seafarers' experience and simplified decision rules, often fall short in complex scenarios. The advent of artificial intelligence, particularly Deep Reinforcement Learning (DRL), offers a promising alternative by enabling systems to learn and optimize collision avoidance strategies through environmental interaction. This study delves into the key technologies behind DEL-based ship collision avoidance, aiming to bolster ships' autonomous avoidance capabilities under a variety of conditions. We begin by outlining the collision avoidance challenge and the fundamentals of DRL, then introduce a novel DRL-based framework designed for this purpose. Utilizing advanced deep neural networks trained on simulated maritime scenarios, our model learns to navigate and avoid collisions in complex situations, outperforming traditional and other AI-based methods in our tests. Notably, the framework demonstrates exceptional performance in multi-ship encounters and constrained waters, showcasing robust generalization abilities to novel environments. By proving DRL's viability in enhancing navigational safety, this work paves the way for future innovations in intelligent shipping, promising significant contributions to maritime safety technologies and intelligent system development. We anticipate that these findings will guide future research and practical implementations, marking a step forward in the pursuit of safer, more autonomous maritime navigation.