Facial reconstruction

Search LJMU Research Online

Browse Repository | Browse E-Theses

A knowledge-free path planning approach for smart ships based on reinforcement learning

Chen, C, Chen, XQ, Ma, F, Zeng, XJ and Wang, J (2019) A knowledge-free path planning approach for smart ships based on reinforcement learning. Ocean Engineering, 189. ISSN 0029-8018

[img] Text
Accepted version.pdf - Accepted Version
Restricted to Repository staff only until 27 August 2020.
Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (983kB)

Abstract

With the development of artificial intelligence, intelligent and unmanned driving has received extensive attention. Compared with the rapid technological advance of unmanned vehicles, the research on unmanned ship technology is relatively rare. The autonomous navigation of cargo ships needs to meet their huge inertia and obey existing complex rules. Therefore, the requirements for smart ships are much higher than those for unmanned vehicles. A smart ship has to realise autonomous driving instead of manual operation, which consists of path planning and controlling. Toward to this goal, this research proposes a path planning and manipulating approach based on Qlearning, which can drive a cargo ship by itself without requiring any input from human experiences or guidance rules. At the very beginning, a ship is modelled in a simulation waterway. Then, a number of simple rules of navigation are introduced and regularized as rewards or punishments, which are used to judge the performance, or manipulation decisions of the ship. Subsequently, Q-learning is introduced to learn the action–reward model and the learning outcome is used to manipulate the ship’s movement. By chasing higher reward values, the ship can find an appropriate path or navigation strategies by itself. After a sufficient number of rounds of training, a convincing path and manipulating strategies will likely be produced. By comparing the proposed approach with the existing Rapid-exploring Random Tree (RRT) and the Artificial Potential Field A* methods, it is shown that this approach is more effective in self-learning and continuous optimisation, and therefore closer to human manoeuvring.

Item Type: Article
Uncontrolled Keywords: 0905 Civil Engineering, 0911 Maritime Engineering, 0405 Oceanography
Subjects: H Social Sciences > HE Transportation and Communications
Q Science > QA Mathematics > QA75 Electronic computers. Computer science
T Technology > TC Hydraulic engineering. Ocean engineering
Divisions: Maritime & Mechanical Engineering
Publisher: Elsevier
Date Deposited: 11 Oct 2019 09:02
Last Modified: 11 Oct 2019 09:02
DOI or Identification number: 10.1016/j.oceaneng.2019.106299
URI: http://researchonline.ljmu.ac.uk/id/eprint/11539

Actions (login required)

View Item View Item