Autonomous collision avoidance control using deep reinforcement learning for maritime autonomous surface ships

Research output: Contribution to conferencePaperpeer-review

3 Downloads (Pure)

Abstract

The maritime industry has been progressing towards autonomous shipping with the main barrier and scepticism eing on the safety assurance of the next-generation autonomous ships. This study aims to enhance the safety of the autonomous ships by developing an intelligent agent that makes evasive decisions considering the ship domain as a safety zone. The proposed approach is demonstrated by considering the case study of a short sea shipping cargo ship. An intelligent reinforcement learning agent is trained to maoneuvre the investigated ship in restricted sea area. The results of this study verify the agent's ability to make safe evasive decisions and control the autonomous collision avoidance for autonomous ships in known and unknown environments.

Conference

ConferenceThe 4th International Conference on Maritime Autonomous Surface Ships (ICMASS) and The International Maritime and Port Technology and Development Conference
Abbreviated titleICMASS/MTEC
Country/TerritorySingapore
CitySingapore
Period6/04/227/04/22
Internet address

Keywords

  • maritime autonomous surface ships (MASS)
  • autonomous collision avoidance system
  • decision-making
  • risk metrics

Fingerprint

Dive into the research topics of 'Autonomous collision avoidance control using deep reinforcement learning for maritime autonomous surface ships'. Together they form a unique fingerprint.

Cite this