A2I2 underwater survey robot thrives in the deep end

An autonomous underwater robot developed with a view to aid nuclear decommissioning has been tested at a demonstration in Cumbria.


Autonomous Aquatic Inspection and Intervention (A2I2) is the work of a wide-ranging UK collaboration led by Rovco that includes Forth Engineering, D-RisQ, the National Oceanography Centre (NOC), Thales UK and Manchester University. The demonstration, which took place at Forth’s 1.2 million litre tank facility in Maryport, saw the robot successfully move autonomously through the deep-water pond, avoid collisions, locate items on the pond floor and relay images in real-time back to screens nearby on dry land.

Sellafield plutonium a multi-layered problem

The capabilities of A2I2 bring together 3D computer vision, simultaneous location and mapping (SLAM), autonomous path planning and scene understanding using Machine Learning (ML). While the technology has been designed primarily for the storage ponds at nuclear sites such as Sellafield, the ‘Intelligent Data Collection System’ can be integrated onto any subsea vehicle to enhance its capability, according to Rovco.

“The whole challenge of this particular project was removing people from hazardous and dangerous environments,” said Gary Cross, senior robotics engineer at Rovco. “One of the key things is increasing the distance between the operators and the environment they’re working in. And the easiest way to do that is to make the vehicle remote, remotely controlled and remotely operated.

“Some of the challenges you have, of course, is that as you increase the distance between the operator and the tools which they’re using, you have increased latency and other challenges which generally make it very difficult to operate. So, technology such as the advanced perception system and mapping capabilities enables us to use the autonomy to provide the appropriate systems to the pilots who can still be in control in a safe and meaningful way within the pond.”


The demonstration was attended by representatives from Sellafield and other potential end users from a range of industries including oil and gas, marine and renewables. Unlike previous trials, A2I2 was able to showcase an ability to reconstruct complex hazardous environments with no prior knowledge, using only its sensor suite and mapping capabilities.

“It’s a huge advancement from what we’ve seen in the previous drop trials,” said Chris Ballard, robotics and artificial intelligence manager at Sellafield Ltd. “The benefit for us within Sellafield is we need to look at removing operators from these harsh environments. Typically, things are done through tele-operational controls. What we’re seeing demonstrated here is the potential for autonomous processes where we can map live environments. It enables us to see what’s going on in our harsh environments.

“It has also got the collision avoidance systems built in and the communications that we so desperately need from down within the pond. So it’s tackling a lot of different discrete areas for us to utilise within the business.”