Understandable Autonomy for Unmanned Maritime Vehicles
Maritime autonomous vehicles are a rapidly developing technology with applications including scientific surveying, unmanned cargo ships and various defence and surveillance roles. For surface vessels there are long established rules governing navigation at sea, the collision regulations or COLREGS, which must be complied with. These rules are easily understandable as well as being sufficiently flexible and general to cover a broad range of possible scenarios and have evolved over time to minimise the risk of collision. Any international regulations governing autonomous ships are therefore almost certain to require compliance with the COLREGS. More generally, there is likely to be a requirement on autonomous systems that they be able to explain any decision which potentially impacts on other vessels. However, in a complex maritime environment developing effective human readable rule-based systems is challenging. This is often a direct result of the constraint of human-readability itself. For example, one characteristic of the COLREGS rules that makes them understandable is that they only refer to pairwise interactions between vessels. However, this is likely to be sub-optimal in a crowded shipping lane, leading to inefficient movement at best and dangerously inconsistent actions, at worst.
Working with an industrial partner this PhD project will investigate the effectiveness of human understandable rules for autonomous navigation and collision avoidance.
Contact: firstname.lastname@example.org for further information.