Decentralised collision avoidance in a semi-collaborative multi-agent system

Abstract

A new approach is considered which explores a decentralised trajectory optimisation algorithm in partly collaborative multi-agent systems to improve safety and provide reliable collision avoidance for vessels in narrow waterways and the open sea. This research will explore trajectory planning under the hypothesis that not all vessels in an encounter will be able or willing to use the same, proposed system. Planning realistic trajectories, which minimise the need to re-plan, will be achieved by observing the predicted behaviour of uncooperative vessels, based on probabilistic models derived from historic data. © 2013 Springer-Verlag Berlin Heidelberg.