IEEE Robotics & Automation Magazine - September 2023 - 12

orchestrates a network of multiple simulators using a fixed or
variable step size to enhance computation speed. It also often
comes with the price of compromised accuracy. Another
important aspect of the component models is standardization
of their interfaces. In many applications, there is a need to
design, simulate, and execute a network of components from
different users. In this regard, the SSP is an extension of the
FMI standard to define complete systems consisting of one
or more FMUs, including their models, with parameterization
and solvers that can be transferred between simulation
tools. The standard and specification enable faster construction
of cosimulation simulators by simplifying the modelconnection
process.
Table 2 includes a collection of components that represent
the Gunnerus digital twin. The components were developed
by many different project partners using various methods and
tools. Most of these components are open sourced and support
the FMI standard, with the exception that a license is needed
for the Sintef models. These components are used as reference
models for the Gunnerus digital twin in the use cases presented
in the " Gunnerus Twin-Ship Validation " section.
TABLE 2. REFERENCE MODELS FOR THE GUNNERUS TWIN
SHIP (ALL DETAILS COULD BE ALSO DOWNLOADED FROM
OSP WEBPAGE [13]).
SUBSYSTEM COMPONENT
Vessel
Hull
Thrusters; include
PMAzimuth thrusters
and bow thruster
Power plant
Thruster drive
Deck
machinery
A frame (plus
controller)
Palfinger crane
(plus controller)
Environment Wind
Current
Global navigation
satellite systems sensor
VRU sensor
Gyro sensor
Controllers DP controller
Waypoint provider
2D/3D (reference
target points)
Observer (position
feedback)
Trajectory controller
TOOL/CODE PROVIDERS
Sintef
Vesim
Vesim
20-sim
20-sim
20-sim
Simulink
20-sim
Winch (plus controller) 20-sim
SimulationX
Sintef
Kongsberg
Sintef
Sintef
NTNU
NTNU
NTNU
Simulink Kongsberg
Simulink Kongsberg
Kongsberg
C++
C++
C++
Java
Java
Java
Java
Kongsberg
Kongsberg
NTNU
Kongsberg
NTNU
NTNU
NTNU
DP: dynamic position; NTNU: Norwegian University of Science and
Technology; VRU: Vertical Reference Unit.
It is worth mentioning that strong-coupled systems are
not naturally easily handled separately in a cosimulation.
As an example, the dynamical inertial impacts between the
ship hull and the deck crane with a heavy payload cannot be
neglected during operation [14]. In this very specific example,
the payload and the crane can be connected via a flexible
cable and treated as one component. The connection between
the ship hull and the crane base can also be represented by
a spring-damper system with extreme high stiffness and low
damping coefficient. However, it requires extreme high computing
resource for real-time simulation, which is critical for
crane operations in this application. Alternatively, the forces
from the crane, including the payload, can also be approximated
as external forces applied to variant attacking points
on the ship hull during operation.
ENABLING TOOLS FOR ONBOARD SUPPORT
Figure 4 illustrates a scheme of enabling tools for onboard support
in our digital twin system. First, the digital twin can store
millions of data points from the physical vessel for different
types of operations. A meaningful usage is model-based sensitivity
analysis for modeling the relevant data and computing how
much the model inputs contribute to its output. Developing such
a supporting tool makes sense, for example, for energy consumption,
as one can analyze what the main factor is during
operation and take alternative solutions to save consumption.
Note that the data sensitivity analysis module is the basis of the
scheme. It takes the ship status, operational commands, and
environmental data as inputs and the designated metric, e.g., ship
position, as the output to quantify how much the inputs contribute
to the output. The result can benefit both the optimization
and the prediction phase.
Second, the digital twin plays the role of modeling and
simulation in the virtual world before executing any onsite
operations for a comprehensive understanding of certain specific
operations, such as mooring, dynamic positioning, and
crane lifting. Dynamic optimization refers to the state of the
ship and the mission being executed. It considers constraints
and generates optimized references for control; meanwhile, it
formulates the references as prior knowledge for predictions of
future operation as far as the control couples with the optimization
module. Similar couplings exist between the prediction
and control modules as they are, in essence, a complete, closedloop
system. In this part, focus is placed on the creation of
valid models by either applying mathematical models or using
data-driven approaches in marine operations. An example is
the parameter identification of a ship model. Another modeling
example is to create a ship motion predictor. The solution could
be a hybrid modeling method that combines a mathematical
model with a data-driven one to estimate a ship's position and
heading in a time series.
The " performance optimization " module plays a role in
optimizing the instantaneous and transient control of a ship
for efficiency or performance, making informed decisions
regarding performance versus time, cost, physical constraints,
and maritime regulations. Path planning is one of the cases on
12 IEEE ROBOTICS & AUTOMATION MAGAZINE SEPTEMBER 2023
IEEE Robotics & Automation Magazine - September 2023

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2023
