J.R. Marx et al. 
Icean Engineering 343 (2026) 123388 
PO 
vr 
YIplB L-uH 
Bene de 
$EFIL' 
A lg LA AP 
.—.x;‚,C 
2 
ig. 1. The maneuvering vehicles: MESSIN. DENEB and BELA 
design and the trajectory optimization for collision avoidance. In each 
of these two sections, the related works will be specifically detailed for 
che topic. A model predictive controller (MPC) is used on all three vehi- 
cles which can switch between different operation modes. Section 3 also 
depiets the underlying motion models, special features in the respective 
allocation and implemented controller modes. The collision avoidance 
algorithm is based on an initial trajectory for each vehicle and an op- 
timization problem that is minimized using defined criteria. Section 5 
presents the developed test procedures, the results, in simulation and 
real-world, and discusses them. In Section 6, the conclusions and future 
works follow. 
Taking into account, in particular, the comprehensive system solu- 
(ons for highly automated or autonomous shipping, one of the most am- 
oitious projects is the major Japanese initiative MEGURI2040 in which 
ooth companies and universities are involved. The project is motivated 
by the aging Japaneses society and the need of ferry connections to the 
nore than 400 inhabited island. The initiative is primarily intended to 
ransfer conventional ships that are already sailing to automatic or au- 
:onomous operation, so that 50 % of coastal ships will be autonomous by 
2040. The Japanese ship classification society ClassNK published con- 
cepts and guidelines for this transfer (Suzuki, 2021; Miyoshi and Ioki, 
2021; ClassNK, 2020). Furthermore, first results of simulations but also 
of actual ship experiments were presented. An automatic collision avoid- 
ance (CA) system was used in Hashimoto et al. (2021) for actual ship 
:ests with the 50 m long training ship Fukae Maru in congested waters of 
Osaka Bay. An ordinary autopilot was combined with a reinforcement 
earning (artificial intelligence - AI) system to control the heading angle 
»ased on the sensors available onboard. Safety criteria were defined as 
ollows: wind speed less than 10 m/s, wave height less than 2 m, visibil- 
ty of at least 2 nm, no abnormalities of nautical instruments, machinery 
ır AI system, and the judgment of captain or watch keeping officer on 
suitable test conditions. In Higo et al. (2023) and Higaki et al. (2024) 
‘he consortium released further results concerning automatic berthing 
and unberthing using neural networks (NN) for trajectory planning and 
racking control with the 222.5 m long ferry Soleil.