DP-Ship

This co-simulation demonstrator is based in the work presented in [1,2], and the reader is referred to these references and the references within for more details. Hence, any in-depth details regarding the submodels are considered out of scope here, except information useful for running the co-simulation demonstrator.

The demonstrator includes five different fmus, and a short description of these are given in Table 1.

Table 1: List of FMUs in the DP-vessel demonstrator case.

Note that all these fmus contains binaries for both win64 and linux64.

Model Descriptions

In the following, each model in Table 1 is presented with focus on running the demonstrator.

OSOM

This fmu contains a simplified vessel model including an irregular sea-state, currents and simple hydrodynamics based on potential theory (including second order wave forces) [4]. The vessel model also contains two rotatable main propulsors placed symmetrical at the stern and one tunnel thruster in the bow.

This model has been exported from 20Sim and because of a limitation in the exporter, the model parameters cannot be changed by the user in a co-simulation. Some of the main parameters that are hard-coded are given in Table 2. The model I/O for the OSOM fmu are described in Table 3.

Table 2: Main vessel parameters.

Table 3: Vessel model inputs and outputs (I/O).

NLPobserver

This fmu contains a wave filter (a Non-Linear Passive observer [3]). The purpose of this filter is to filter out the fastest oscillating wave-induced motion disturbances from the position and orientation measurements from the vessel, the ones that the onboard power and propulsion system cannot compensate for without drawing an enormous amount of power. The filter also estimates the position and orientation rates which the DP-controller needs for providing sufficient damping to the vessel motions. Note that also this fmu is generated from exported code from 20Sim, hence, no scenario parameters can be changed by the user. Nevertheless, the parameters hard-coded in the NLPO match the parameters in the vessel model.

The model I/O for the NLPobserver is given in Table 4.

Table 4: Wave filter inputs and outputs (I/O).

ReferenceGenerator

This fmu contains a reference position generator, which, based on user defined inputs, generates continuous reference signals for the DP-controller (north, east and yaw). This fmu is implemented in C++ and the parameters are accessable for the user and can be changed. The most important parameters are listed and described in Table 5.

Table 5: Main parameters and I/Os in ReferenceGenerator.

To illustrate how the ReferenceGenerator works, consider the following:
At co-simulation time t0 the vessel heading is psi0 and the user sets

psi_wp = psi0 + pi/4
psi_tp = t0 + 50.0

Then, as long as

pi/200.0 <= dpsiMax

The rate of the yaw-reference signal is set to pi/200. If not, it is set to dpsiMax. Next, this yaw-rate gets integrated and smoothed using low-pass filter (with filter time constant Tpsi). Note that also the signal rate (dpsi_ref) gets smoothed using a low-pass filter. Afterwards, both dPsi_ref and psi_ref are fed to the DP-controller.

DPController

This fmu contains the DP-controller which controls the position and the orientation of the vessel. Also this fmu is implemented in C++ and the parameters (controller tuning gains) are accessable for the user and can be changed. However, the controller has been tuned and there is no need for changing these paremeters and it is recommended that these parameters are left as is. Table 6 lists the main paramters and I/Os in the fmu. The control law itself is a simple PID-based control law including the rotational matrix for the vessel orientation (yaw).

Table 6: Main parameters and I/Os in DPController.

ThMPC

This fmu contains the thrust allocation algorithm that takes the commanded global thrust forces from the DP-controller and distributes them as local thruster signals in type of thruster orientation and thrust amplitude. This thrust allocation algorithm is based on Model Predictive Control-theory (MPC) and is thoroughly presented in [5]. Hence, no details will be given here. The main parameters and I/Os in this fmu is given in Table 7.

Table 7: Main parameters and I/Os in ThMPC.

Note that there also exists other parameters and variables, but these are the most important. To see the others the readers are referred to the ModelDescription.xml-file in the fmu. Also note that the ThMPC-fmu only runs the MPC solver routine every second, and uses the obtained thrust-rates to calculate thruster commands inbetween these quadratic problem (QP) solver time steps.

Model Connections

In this demonstrator case, the model connections are given as follows:

OSOM.Thrust_d[1]        <-  ThMPC.F1c
OSOM.Thrust_d[2]        <-  ThMPC.F2c
OSOM.Thrust_d[3]        <-  ThMPC.F3ci

OSOM.input[1]           <-  ThMPC.alpha1
OSOM.input[2]           <-  ThMPC.alpha2

DPController.x          <-  NLPobserver.y_hat[1]
DPController.y          <-  NLPobserver.y_hat[2]
DPController.psi        <-  NLPobserver.y_hat[3]

DPController.dx         <-  NLPobserver.v_hat[1]
DPController.dy         <-  NLPobserver.v_hat[2]
DPController.dpsi       <-  NLPobserver.v_hat[3]

DPController.x_ref      <-  ReferenceGenerator.x_ref
DPController.y_ref      <-  ReferenceGenerator.y_ref
DPController.psi_ref    <- ReferenceGenerator.psi_ref

DPController.dx_ref     <-  ReferenceGenerator.dx_ref
DPController.dy_ref     <-  ReferenceGenerator.dy_ref
DPController.dpsi_ref   <- ReferenceGenerator.dpsi_ref

NLPobserver.y[1]        <-  OSOM.q[1]
NLPobserver.y[2]        <-  OSOM.q[2]
NLPobserver.y[3]        <-  OSOM.q[3]

NLPobserver.tau[1]      <-  ThMPC.Fxg
NLPobserver.tau[2]      <-  ThMPC.Fyg
NLPobserver.tau[3]      <-  ThMPC.Mzg

ThMPC.refx              <-  DPController.Controlx
ThMPC.refy              <-  DPController.Controly
ThMPC.refMz             <-  DPController.ControlMz

More information on how to use the cosim demo app is available in the user guide, where the DP-Ship demo test case was used as an example.

References and Footnotes

[1] S. Skjong, M. Rindarøy, L. T. Kyllingstad, V. Æsøy, and E. Pedersen, “Virtual prototyping of maritime systems and operations: applications of distributed co-simulations,” J. Mar. Sci. Technol., pp. 1–19, 2017.
[2] S. Skjong and E. Pedersen, “Co-Simulation of a Marine Offshore Vessel in DP-Operations including Hardware-In-the-Loop (HIL),” in Proceedings of the ASME 2017 36th International Conference on Ocean, Offshore, and Arctic Engineering OMAE 2017, 2017.
[3] T. I. Fossen and J. P. Strand, “Passive nonlinear observer design for ships using lyapunov methods: full-scale experiments with a supply vessel,” Automatica, vol. 35, no. 1, pp. 3–16, Jan. 1999.
[4] O. Faltinsen, Sea Loads on Ships and Offshore Structures. Cambridge University Press, 1993. [5] S. Skjong and E. Pedersen, “Non-angular MPC-based Thrust Allocation Algorithm for Marine Vessels -A Study of Optimal Thruster Commands.”

¹Note that q is used here as name for the vector containing the vessel position and orientation. This has to do with the name convensions in bond graph modelling theory, where q is often used for the state vector containing position and orientations in mechanical systems.