CONTROL METHOD FOR PATH FOLLOWING AND COLLISION AVOIDANCE OF AUTONOMOUS SHIP BASED ON DEEP REINFORCEMENT LEARNING
Autonomous ships have received increasing attention in the maritime industry. The development of a real-time path following and collision avoidance system complying with the International Regulations for Preventing Collisions at Sea (COLREGs) is crucial to the development of autonomous ships. In this study, we proposed a novel deep reinforcement learning (RL) algorithm to improve the efficacy and efficiency of the path following and collision avoidance system. To verify the proposed algorithm, we conducted simulations of an autonomous ship under unknown environmental disturbances to adjust its heading in real time. A three-degree-of-freedom dynamic model for the autonomous ship was developed, and a line-of-sight (LOS) guidance system was used to guide the autonomous ship along a predefined path. Then, a proximal policy optimization (PPO) algorithm was implemented for the problem. We found that, after applying the advanced deep-RL method, an autonomous ship could learn the safest and most economical avoidance behavior through repeated trials. The simulation results showed that the proposed algorithm guaranteed collision avoidance with encountered moving ships while ensuring that the ship followed a predefined path. Additionally, the algorithm demonstrated that it could manage complex scenarios with various encountered ships in compliance with COLREGs, showing excellent adaptability to unknown complex environments.
Zhao, Luman; Roh, Myung-Il; and Lee, Sung-Jun
"CONTROL METHOD FOR PATH FOLLOWING AND COLLISION AVOIDANCE OF AUTONOMOUS SHIP BASED ON DEEP REINFORCEMENT LEARNING,"
Journal of Marine Science and Technology: Vol. 27:
4, Article 1.
Available at: https://jmstt.ntou.edu.tw/journal/vol27/iss4/1