Particle Swarm Optimization for Dynamic Risk-Aware Path Following for Autonomous Ships

Abstract

This work presents the use of particle swarm optimization (PSO) for dynamic risk-aware path following, or waypoint re-planning during autonomous surface navigation in a maritime environment with polygonal grounding obstacles. Although recent research on control and local or global path planning for maritime autonomous surface ships (MASS) is considerable, few deal with the concept of risk during dynamic path following along a preplanned path. The proposed method introduces risk-based terms in the PSO fitness function for dynamic (online) adjustment or re-planning of intermediate waypoints during path following of a pre-planned route, where the control of the vessel in this work is left to a standard line-of-sight PID controller. Moreover, the results are compared to the performance of an analogous implementation of risk-aware model predictive control (MPC) using a gradient-based solver. The suggested method yields adequate performance similar to that of the MPC algorithm. Ultimately, the PSO approach may in future works allow for more complex and dynamic risk-aware behavior related to e.g. weather conditions implemented through lookup tables, or discrete machinery modes that are non-smooth.