Verification methodologies for autonomous agents to enable safe marine autonomous surface ships (MASS)

The application of autonomy is driven by a desire to safeguard lives by removing crew from dangerous environments and reducing the likelihood or impact of accidents arising from human error. State of the art Marine Autonomous Surface Ships (MASS) deployed for extended periods in complex in-land, coastal, and off-shore applications, must be able to address changes in their state, environment, and mission by adapting their mission plan. While ongoing developments in robotic autonomy result in systems that are capable of creating adaptable mission plans, due to the complexity, the plans these systems propose may not be predictable. It is therefore important to develop and implement verification methodologies to minimize the risk of incorporating these autonomously-created mission plans, and build confidence in their use.

This project will develop a novel verification methodology building from an understanding of traditional and emerging off-line and run-time verification methods, as they apply to autonomous mission planners that cannot be verified through existing codes and standards. The ultimate objective is to enhance the safety of the MASS and other entities (e.g. people, ships, other unmanned systems etc.) it interacts with and those it is not intended to interact with (like marine mammals, the seabed etc.).

This research brings together complementary capabilities of Dalhousie University in research and development of autonomous systems, and the expertise of Martec Ltd. in compliance and verification of manned and unmanned ships. The project will help extend Martec’s capabilities in autonomous marine systems to provide access to this active market segment.

Faculty Supervisor:

Mae Seto

Student:

Amy Deeb

Partner:

Martec Ltd

Discipline:

Engineering - mechanical

Sector:

Professional, scientific and technical services

University:

Dalhousie University