Collision Avoidance for ASVs Using Model Predictive Control
Master thesis
Permanent lenke
http://hdl.handle.net/11250/2433779Utgivelsesdato
2017Metadata
Vis full innførselSamlinger
Sammendrag
If an Autonomous Surface Vehicle (ASV) is to be operated at sea in the proximity of other vessels it must be equipped with a Collision Avoidance System (CAS). This system must comply with the rules for avoiding collision that govern the seas, the COLREGS.
In this thesis a COLREG compliant CAS using Simulation-Based Model Predictive Control (SB-MPC) has been implemented. The four main scenarios used to test the method's behavior and COLREGS compliance is: head-on, overtaking, and crossing from port and starboard. Several more complicated scenarios are also studied to see how the method deals with more complicated situations. The tests were performed within a Robotic Operating System (ROS) framework using a non-linear ship model to predict the \gls{asv}'s movements. The tests were then repeated using a linear model. The behavior of the system is discussed and compared with the behavior of the ASV equipped with a CAS based on the Velocity Obstacle (VO) algorithm.
After the simulation study was completed the CAS was also tested on the On-board System Simulator (OBS) of Maritime Robotics to prepare for full-scale experiments. These simulations were using live Automatic Identification (AIS) data from ships in the Trondheimsfjorden.
The CAS performed well during the simulations, both within the ROS framework and in the OBS environment. However, finer tuning is still needed to get optimal performance.
Full-scale testing was also attempted, but not completed, mostly due to the lack of suitable obstacle vessels.