Where humans fear to tread: autonomous robots in hazardous environments

Just two hours: that’s the amount of time some robots lasted that were sent into the devastated Fukushima nuclear power plant, due to the crippling levels of radiation – and that was in 2017, six years after the initial accident.

Now as the UK’s 15 nuclear reactors and another 400+ around the world face astonishingly difficult decommissioning and scheduled maintenance operations, the need for intelligent autonomous robots to enter hazardous spaces and perform effectively has never been greater.

Developing computer programmes

Scientists at the University of Liverpool are at the vanguard of this effort to develop better computer programs for better autonomous machines, so that every second in such hazardous places counts. “We don’t want people to go into these nasty areas, we want robots to go there instead,” says Professor Michael Fisher from the Department of Computer Science. “We need to develop robots for roles in nuclear decommissioning, such as monitoring, reconnaissance and deconstruction.”

When robots are in these hazardous environments, it's essential that they make the right decisions. Fisher and colleagues are tasked with generating, testing, and validating the computer code to make sure these decisions happen correctly every time.

It’s a significant verification task to make sure that when a robot is built, all of its components are doing what they should be. “Distinct software has to be made to make key decisions, possibly buried in millions of lines of other code, and analysing that software is the core of what we’re involved in,” says Fisher who is also Director of the Centre for Autonomous Systems Technology (CAST) at Liverpool. “Then we make mathematical proofs to ensure the robots will always uphold those decisions, followed by tests in real world environments.”

Research applications

The nuclear energy industry is one of the areas where huge investments are being made developing the artificial intelligence of autonomous and semi-autonomous systems. The others are in space, the oceans and the domestic home.

The work is supported by the UK Industrial Strategy Challenge Fund (robotics and AI category) where more than £93 million is available from 2017-2021 to strengthen UK science and business innovation.

Working in Partnership

The University of Liverpool is a partner in three of four major research hubs, involving other UK universities and various industrial partners, that have been awarded £44 million to develop robotics and AI for use in extreme environments.

The nuclear RAIN Hub (Robotics and Artificial Intelligence for Nuclear) is led by the University of Manchester and, as well as Liverpool, involves collaborators from the Universities of Bristol, Lancaster, Nottingham, Sheffield and Oxford. It also includes partners such as the UKAEA’s Remote Applications in Challenging Environments (RACE) centre, major industry players Sellafield Ltd and EDF Energy, other international partners in robotics and nuclear engineering from the US, Japan, and Italy, as well as a clutch of SMEs across the UK.

Fisher says that the University of Liverpool is ideally placed to deliver on these modern challenges in applying cutting-edge science for positive societal benefit. “At Liverpool the research that goes on is world leading in the areas we need, such as robotics, AI and the computer sciences.” Other national capabilities are drawn upon too, such as the UK’s supercomputers and simulation facilities at the Virtual Engineering Centre at Hartree.

Back to: Research