Improving subsea navigation using environment observations for long-term autonomy

Precise navigation is a fundamental requirement for robotic competence. However, Global Navigation Satellite Systems (GNSS) such as GPS are unavailable in water, and current navigation solutions rely on separate ship or seafloor instrumentation that dominates operational costs and limit efficiency gain. 

You will investigate how Simultaneous Localisation and Mapping (SLAM) techniques can minimise navigation errors by taking advantage of visual and/or terrain-based observations of the environment, matching these to elements stored in self-generated or pre-loaded maps.

Although modern SLAM solutions are satisfactory for autonomous robotic missions that cover ~100km trajectories and last days, they become computationally unsolvable as survey missions grow in duration and extent.

With advances in high-density energy storage and efficient propulsion systems, missions covering 1000km+ and lasting several months are becoming mechanically possible. We need better SLAM frameworks to match the mechanical developments and identify:

• minimal navigation setups for persistent, precise navigation 
• impact of adding observation sources on precision and computational complexity

In this industry-funded PhD, you will have access to data from over 100 past AUV missions. You will develop both simulations and perform practical experiments with state-of-the-art AUV platforms and experimental facilities at UoS and Sonardyne International Ltd, and there will be opportunities to participate in fieldwork deploying robotic systems at sea during your PhD.

You will receive regular support from the supervisory team and work within a group of underwater robotics researchers and benefit from strong collaborative links with leading international marine robotics groups (includes NOC, Southampton; ACFR, USydney; IIS, UTokyo and Carnegie Melon University).

Darryl Newborough at Sonardyne International Ltd will form part of the supervisory team.