Model-based Control of Small-scale Surface Vessel

Introduction
The purpose of the project was to investigate automation of marine vessels. This was done using a small-scale surface vessel previously developed bu several master thesis projects (latest by G. Zetterqvist and F. Steen, see link below). With aid from the industry in the form of an expert from ABB, the project is lead by Linköping univeristy.
G. Zetterqvist & F. Steen

Background
A cornerstone of the development of the shipping industry is automation. Autonomous ships can save time, money and increase both reliability environmental sustainability. In order to develop these systems, efficient methods for estiamting model parameters are crucial and model based control can make good use of these models.
Increase reliability
Increase sustainability
Improve efficiency

Goals
The goals of the project were to evaluate an previously presented model and streamline data collection for parameter estimation. Develop a model based controller and build a GUI to aid in the future development of the vessel systems.
Model evaluation
Controller implementation
Graphical User Interface

See the vessel in action and get familiar with the features of the system.

Model
The previously developed boat model has three degrees of freedom. Apart from the states and inputs it has a number of unknown, time-static, parameters. The model is implemented in MATLAB as a nlgrey-object and the unknown parameters are estimated using nlgreyest .
The states of the model are shown below. The signals that can be used to control the real boat are the model's input signals.
The purpose of the model is to be able to predict the vessels movement for a given time horizon. Therefore, the performance is validated by comparing the model's predicted values of the states with the measured states, see the plot below for predicted vs measured position:

Controllers
The implemented controllers both follow a predefined reference path and the objective is to make the vessel follow both position and heading as closely as possible.
The MPC (Model Predictive Control) uses a linearized approximation of the previously developed model. The model uses errors relative to the reference as states. The model is used to predict the movement of the vessel and a quadratic programming optimization problem is solved find the best control-signal for the vessel.

As seen in the figure above, there is still performance to extract. This can be achieved by fine tuning the parameters of the algorithm. It is also possible to extend the model to include limitations or weights on the rate of change of the control-signals.
The SFC (State Feedback Controller) is sufficiently compliant with the given reference. It is derived from a single-track bicycle model and operates with continuous rpm on the azimuth thrusters. Only the angles of the thrusters are controlled.

GUI
The graphical user interface can specify waypoints for a desired route, designate and fine tune a controller and track the vessel in real time.
It is built in Matlab and communicates with the vessel via ROS.

Path planning
The user specifies waypoints through the GUI. The system then interpolates using a spline function.
See the image below for a demonstration of the interpolation functionality
The waypoints contains information regarding position, heading and speed, the interpolated path is then utilized as a reference by the vessel.

Below you will find all documents created in this project.
Note that most of them are written in Swedish.

       Project plan
    
       Design specification
    
       Requirement specification
    
       Poster

       Test plan
    
       Test protocol
    
       User manual
    
       Technical report

This is the group of people behind the project

Oscar HermanssonDocumentation
Oskar JonssonHardware
Jonatan SiönäsDesign

Axel StåhlbomSoftware
Johannes WenngrenProject manager
Tim WiikTesting

Below you will find a selections of interesting photos taken during the project.