/.R. Marx et al. 
„A 6sier €. 
Supervisor 
Switch 
Control 
Mode 
Sens 
‚NE WaLe 1616 
m 
Actuators 
DA 
AS 
Ar, 
INC 
Initial Maneuver Plan & Current States 
witch 
;ontrol 
"-de 
Achıators | 
it), 
. 
{ 
a 
ar 
I BA 
18164. 
Icean Engineering 343 (2026) 123388 
Switch 
Zontrof 
"ode 
Actuators 
ul 
dE 
r;R 
mE TE AA 
In hoard DFM- 
Fig. 2. Overview of the vehicles and general data connections. 
py disturbances caused by wind and currents, which can be taken into 
account by the MPC (Marx et al., 2024). 
The A-SWARM project has also already tested the high-level automa- 
(on tasks path and motion planning as well as automatic collision avoid- 
ance in real experiments (Damerius et al., 2024). Detailed publications 
[or a real-time path planning method and the fast generation of a fea- 
sible trajectory can be found in Damerius and Jeinsch (2022) and in 
Damerius et al. (2023). 
For efficient development in the numerous maritime automation 
projects at our institute, an automation system was created as a gen- 
aral basis that can be easily adapted to the respective vehicle. It repre- 
sents the modules for guidance, navigation and control (GNC) and pro- 
vides the interfaces to the ship’s sensors and actuators. Particularly when 
‘he vehicles are not owned by the university, rapid deployment and re- 
noval of the automation system is essential for the short testing periods. 
Overviews to this system concept were given in Kurowski et al. (2019a) 
[or the first tests in project GALILEOnautic, in Schubert et al. (2023) for 
different vehicle solutions and in Damerius et al. (2024) specifically for 
the A-SWARM approache. 
2. Test design 
The section on the test design begins with a description of the vehi- 
cles involved, their equipment and the communication between them. In 
a further subsection, the design of the maneuver scenario is explained. 
A third subsection is focusing on nautical relevant safety aspects. Fi- 
nally, the initial scenario is presented, which results from the above re- 
]Juirements and which must be resolved collision-free during automatic 
maneuvering. 
2.1. Vehicles, equipment and networking 
The three involved vehicles are seen in Fig. 2, the research ves- 
sel DENEB and the catamarans BELA and USV MESSIN, embedded in 
che communication structure of the general scenario. The specifica- 
tions of the vehicles are listed in the Table 1 with the hull dimen- 
sions, maximum speeds, actuators and standard sensor configurations. 
Due to their actuator equipment, DENEB and BELA are fully actuated 
and therefore capable of DP, MESSIN is not. The vehicles have already 
veen used for automatic maneuvering with different control approaches 
and sensor equipment reported in Section 1.2. While DENEB and BELA 
have their own bridge, the MESSIN can be controlled manually via a 
remote control device (RCD) from land or from an escort vehicle. 
Each of them is equipped with its own automation system with 
the corresponding modules for guidance, navigation and control (GNC) 
which are marked in petrol in the figure. Cooperative maneuvering uses 
ı central guidance module installed on a PC physically located on board 
‘he DENEB. This central system is connected with the three vehicles 
via VPN or LAN to send trajectories and receive current motion states. 
3ased on the initial, offline planned maneuver trajectories, optimal eva- 
sion trajectories are calculated for the cooperative scenario, as described 
.n more detail in Section 4. 
For every navigation solution, the sensor data is fused to suit the 
respective sensor configuration. These fusions rely substantially on the 
output of the high-quality inertial navigation systems (INS) installed in 
each vehicle and are supported by motion filters (Schubert et al., 2023). 
An additional inclusion of lidar data for near-field detection of fixed and 
noving objects and for verification of the map data will be depieted in 
‚urther publication to this project. 
As the automatic identification system (AIS) data is only transmitted 
at a low frequency, especially at low vehicle speeds, a generic inter- 
nodule protocol is used for communication between the vehicles. This 
protocol was developed to exchange generic messages between the GNC 
nodules regardless of the application and vehicle type. In addition to 
‘he status variables, it also contains the information on vehicle name 
ınd size that is essential for AIS data. 
The controller, an adaptive MPC for the tests presented here, can di- 
‘ectly control the vehicle when switched to automatic mode. For the pre- 
sented results of cooperative automatic maneuvering only MPCs were 
applied described in detail in Section 3. 
he extra note that the captain of the DENEB is also the supervi- 
sor of the entire maneuver illustrates the clear hierarchical structure of 
‚esponsibility in these tests. If the captain judges a situation to be too 
:isky or detects too great a deviation from the original plan in the mo- 
:ion states of the vehicles, it is up to him to abort the test and take over 
he manual control. To make it easier to assess the situation, the three 
original maneuver plans respectively the calculated evasion trajectories 
and the current vehicle poses in the MAS are displayed compactly in 
‘he electronic navigational chart (ENC), as seen in Fig. 3. The MAS is 
nstalled above the bridge consoles to be clearly visible. As with previ- 
us automatic tests with the DENEB alone, the motion prediction, the 
quantified motion states and manipulated variables are also displayed 
ın the MAS (Rethfeldt et al., 2021). 
In addition to the MAS on the DENEB, there is a visual system for the 
vehicle operators on each vehicle or the escort vehicle for the MESSIN 
as part of the automation system. It displays the evasion trajectories.