ABSTRACT
We discuss potential assurance frameworks for autonomous navigation systems in the maritime industry, with emphasis on testing and verification of the system’s perception performance and capacities. Ongoing research in this field has revealed profound challenges related to artificial situation awareness and machine perception specific to the marine environment. The lack of a clear and transparent framework and methodologies to assure the safety associated with the usage of such solutions, have been identified as key barriers for the implementation of autonomous navigation solutions at scale. Because the machine perception and situational awareness algorithms are expected to be partly or fully based on machine learning algorithms, including deep learning, whose functional reasoning is challenging or even impossible to understand and predict, the verification of such systems is fundamentally different from a traditional verification process based on physical understanding and theory. We review several methods for testing autonomous navigation systems, proposed and used mainly in the automotive industry, and discuss how these methods can be adapted, combined and applied to form a framework for assurance of autonomy in the maritime industry.
