PhD defence by Thomas Thuesen Enevoldsen

Title: Informed Sampling-based Collision and Grounding Avoidance for Autonomous Marine Crafts

Supervisors
Principal supervisor: Associate Professor Roberto Galeazzi, Department of Electrical and Photonics Engineering, DTU, Denmark
Co-supervisor: Emeritus Mogens Blanke Department of Electrical and Photonics Engineering, DTU, Denmark

Assessment committee
Associate Professor Evangelos Boukas, Department of Electrical and Photonics Engineering, DTU, Denmark
Associate Professor Wasif Naeem, Queen’s University of Belfast, United Kingdom
Associate Professor Anastasios Lekkas, Norwegian University of Science and Technology, Norway

Master of the Ceremony
Associate Professor, Søren Hansen, Department of Electrical and Photonics Engineering, DTU, Denmark

Abstract
Autonomous mobility and transport at sea is projected to be one of the great disruptors towards a sustainable transformation of our society by increasing energy efficiency and safety of maritime transport, as well as de-congestioning urban areas by enabling novel waterborne mobility solutions in connection with the electrification of city ferries. The pathway towards fully autonomous solutions for the ship navigation is still long and arduous, as current regulatory frameworks can be partially encompassed by actual AI-based technologies. However, the maritime sector could reap the benefits of autonomous technologies to act as on-board decision support systems, in order to aid navigators with situation awareness and collision avoidance.
The collision avoidance system is a key component towards the actualisation of autonomous navigation solution at sea. Such system shall assist the human navigator to resolve possible ship collision and grounding situations by proposing alternative paths that are compliant with the international rules for navigation and optimal with respect to specific navigation criteria. The PhD thesis titled “Informed Sampling-based Collision and Grounding Avoidance for Autonomous Marine Crafts” investigates methods and tools for the design of collision avoidance systems and demonstrate them on case studies relevant for maritime transport and waterborne urban mobility.
The novelty of the research is three-fold. A unified motion planning framework that concomitantly encompasses collision and grounding scenarios in the presence of multiple vessels. A probabilistic model describing the collective navigation experience of seafarers while navigating in confined waters to bias the collision avoidance system towards the selection of human-like solutions. Specialized data-driven techniques for decreasing the time to compute rule-compliant and optimal path deviations.
The proposed algorithms for collision avoidance are part of the Autonomous Navigation System onboard the Greenhopper, which will operate in Limfjorden (Aalborg), and they could be included in future commercial navigation systems for autonomous ships.