Situational Awareness for Autonomous Marine Vessels

A key component for maintaining safety while navigating onboard a vessel, is to obtain and maintain situational awareness. Situational awareness comprises perceiving where you are and where others are in relation to you, and relate this information to knowledge and prior experiences, such that one can make informed decisions, which take the possible future evolution into account.
Within the maritime domain, situation awareness embodies the assessment of collision risk with nearby vessels and objects, and given that a risk exists, the evaluation and application of the correct navigational rule, which are specified in the International Regulations for Preventing Collisions at Sea (COLREGs). COLREGs are determined by the Internal Maritime Organisation (IMO), which is the maritime body of the United Nations (UN), and are known as the “rules of the sea”. It is required, by international law, that any and all vessels follow these rules, as failure to do so can have disastrous consequences, e.g. collisions or groundings can cause the loss of human lives, and possibly irreparable damage to the environment. The aim of this thesis is to develop a COLREGs compliant framework for situational awareness for Marine Autonomous Surface Ships (MASS). For this purpose, a mathematical model of the navigational process of human navigators is formalised in Chapter 4, and combined with the theory of Discrete Event Systems (DES), an automata based solution is developed to generated an algorithmic implementation of the human navigational process. The inherent symmetry in the COLREGs obligations between two vessels is exploited in Chapter 5 to extend the situational awareness to include multi-vessel situations, with a complete understanding of the obligations of all involved vessels, an essential step for correctly predicting the future evolution of a multi-vessel encounter. Navigation in confined waters require deeper understanding of the manoeuvrability of other vessels, as certain COLREGs rules might apply, changing the obligations. An method for estimating the manoeuvrability, based on the feasible depth contours for the target vessel, is given in Chapter 6, and is utilised to correctly assess the application of COLREGs Rule 9. Dealing with uncertainty and ambiguous situations is a natural part of navigators lives, so much so that COLREGs specify how theses situations should be resolved. As uncertainty is naturally occurring in any autonomous system which relies upon noisy measurements of its surroundings, thus it is essential for an autonomous system to have the capability to reason about its own uncertainty, so that it may apply the correct COLREGs rules. Chapter 7 explores how the uncertain measurements affects the navigational process, and how it translates into the situation awareness of the system. A requirement for any safety-critical system is to be fault-tolerant, and with more complex software critical to the decision-making, cyber-resiliency is an integral part. A modular software architecture is developed, along with the underlying communication middleware software, which facilitated a decentralised and cyberresilient software framework is developed in Chapter 8.2. The chapter also proposes a autonomous supervisor for decision-making, and the decomposition of this into the three separate entities which were implemented onboard the final testcase. To facilitate the certification of such an autonomous system, each module and their responsibilities were mapped to the standards for which navigators must adhere, known as the International Convention on Standards of Training, Certification and
Watchkeeping for Seafarers (STCW). The thesis consists of nine chapters, which summarises the scientific findings of this research project, which has been disseminated through articles in leading scientific journals and peer-reviewed conference proceedings. These articles are appended to the thesis.