Automatic testing of maritime collision avoidance algorithms

Abstract

For Autonomous Surface Vessels (ASVs) to be put into commercial operation and gain public acceptance, they will have to be as safe, or safer then similar vessels operated by humans. Today, maritime control systems are verified by class societies with established methods, such as hardware-in-the-loop (HIL)-testing. ASVs are equipped with collision avoidance systems and are critical for the safety of ASV operations. Sparse research exist on verification of these systems and their underlying collision avoidance algorithms.

In this thesis a framework for enabling automatic testing of collision avoidance algorithms is developed. Two collision avoidance algorithms are implemented. One is based on the velocity obstacle and the other is based on a type model predictive control, called simulation-based control behavior selection. These systems are tested in automatically generated scenarios and evaluated using a set of metrics. The implemented metrics are developed in the PhD thesis by Kyle Woerner. The implemented metrics are able to detect COLREGS violating behaviors and some characteristics of each system are identified.