Optimization-based Trajectory Planning and Automatic Docking for Autonomous Ferries

Abstract

This Ph.D. thesis presents developments towards energy-optimized trajectory planning and automatic docking for surface vessels. It comprises eight peer-reviewed research papers, in addition to the introductory chapters that provide context to the thesis’ main topics and a summary of the contributions. Autonomy can lead to safer and more energy-efficient maritime operations and enable new, low-cost opportunities within the maritime domain. Commercial actors have already utilized autonomous technology for maritime transportation, surveillance, and research operations. The Norwegian government and several commercial actors are currently exerting efforts to electrify and bring autonomy to car and passenger ferries. Through the Norwegian Research Council’s project “Energy-optimized concept for all-electric, emission-free and autonomous ferries in integrated transport and energy systems,” this thesis focuses on motion control and planning during the three phases of automatic crossing: undocking, transit, and docking. The methods developed in this thesis are focused on practical applications and are tested in full-scale experiments. The thesis contributes to model-based, energy-optimized trajectory planning, a collision-avoidance system compliant with the COLREGs (International Regulations for Preventing Collisions at Sea), a method for automatic docking, and a system for performing automatic crossing, which comprises the phases mentioned above: undocking, transit, and docking. Additionally, the thesis contributes to development and improvements for the experimental autonomous test ferry milliAmpere. As part of this thesis, a method for optimization-based model identification for milliAmpere has been developed. The result is a nonlinear three-degree-of-freedom dynamic model, plus a set of models for thruster dynamics, thruster force mapping, and wind effects. Additionally, the Ph.D. work has contributed to a web-based graphical user interface for milliAmpere and a software-in-the-loop simulator. These contributions have helped lower the threshold for experimental testing of autonomous technology. Most of the academic literature on trajectory planning is towards timeand distance-optimized trajectories, with little focus on energy-optimization and closed-loop results. This thesis presents four model-based, energyoptimized trajectory planning methods for autonomous surface vehicles (ASVs), with various warm-starting techniques, obstacle representations, and solver designs. The most recently proposed method can consider arbitrary maps and disturbances in the form of wind and has been shown to produce dynamically feasible trajectories in validation experiments. In order to consider moving obstacles in accordance with the COLREGs, dynamic collision avoidance methods must recognize various situations when encountering other vessels and produce behaviors that fulfill the rules of those regulations. In this regard, a hierarchical collision-avoidance system that complies with rules 8 and 13–17 of the COLREGs was developed as part of this thesis. The collision-avoidance system is a three-layered architecture that handles nominal, global trajectory planning, long-term dynamic collision-avoidance that considers the COLREGs, and short-term, safe maneuvering. The system has been tested in simulations, where it safely handled a wide range of situations. This work is among the first in academia that complies with the most critical parts of the COLREGs regarding motion planning and control. The literature on automatic docking in the maritime domain is mostly limited to underwater vehicles. The few publications that deal with surface vessels fail to consider obstacles and the harbor layout explicitly. This thesis presents work on an optimization-based method for automatic docking of ASVs, which takes into account harbor layouts and obstacles, using a priori map information and information from exteroceptive sensors. The method is successfully tested in several full-scale experiments. While some commercial actors have performed sea trials with surface ships crossing automatically from one dock to another, they do not publish information about the methods used in such operations. There are also few results from automatic crossing in the academic literature. In this thesis, a framework for automatic crossing for a surface vessel is presented. The framework combines different trajectory planners and control modules to automate the three phases of undocking, transit, and docking. The framework has been tested experimentally and has resulted in successful automatic dock-to-dock crossing operations. Development and improvement of technologies such as trajectory planning, COLREGs-compliant collision avoidance systems, and automated docking methods are small steps towards an autonomous future at sea.