Imagine a military robot that can formidably augment the firepower of an army patrol unit, then switch modes to carry wounded soldiers back to base. It sounds like a scene from Terminator, but such high-tech robotics remain a distant prospect. Robots are not yet capable of reliably seeing, comprehending and assessing what is unfolding around them.
However, a new Defence Cooperative Research Centre (D-CRC) for Trusted Autonomous Systems has recently been formed to help close the gap between the tools and knowledge we currently have and the future technologies we can imagine — innovations that will dramatically extend the capabilities of our defence forces and make our personnel safer than ever before.
The new D-CRC was announced by Defence Minister Christopher Pyne in December 2017, and will start out with $50 million in funding from the Next Generation Technologies Fund.
“The D-CRC’s overarching goal is to deliver world-leading autonomous and robotic technologies to the Australian Defence Force, which will enable trusted and effective co-operation between humans and machines,” says Prof Rob Sale, interim D-CRC CEO.
To start with, projects are being proposed and led by Australian industry, but will also have input at every stage from the Department of Defence Science and Technology (DST).
Australian companies and universities are doing world-class work in many areas relevant to Trusted Autonomous Systems, but they are broadly distributed, says Sale. The D-CRC instead aims to “integrate the talent pool spread across the country and coordinate their efforts,” he adds.
Ethical and legal parameters
Alongside projects developed by teams in the maritime, land and aerospace domains, the D-CRC will run activity groups to explore and mature ethical and legal frameworks for future autonomous systems to operate.
Dr Jason Scholz, chief scientist and engineer for the D-CRC on Trusted Autonomous Systems, says that some of these activities aim to address ethical points, such as “Should we do it?” and “Why would we do it?”. Other activities will consider legal protocols for how people and machines can work together.
For example, Scholz’s team is already investigating a commander–machine legal-agreement protocol in which a human commander might define a goal and a pool of machines then identify which aspects of that goal they can achieve or contribute towards.
“The commander would in effect accept or reject each of those offers, just like we do in contract law, thereby binding the machines for a period of time to do that work,” Scholz says.
Seeing, sensing, perceiving
These are all ideas for the future. “Robots generally lack a ‘theory of mind’. They don’t realise that some objects have beliefs and desires,” says Scholz. The Australian Centre for Robotic Vision is engaged in groundbreaking work to improve the ‘visual’ systems of robots, including sonar detection for underwater robots and infrared vision for robots to operate at night. Seeing is one thing. But what’s next?
“Once we have robots that can see, we want robots to comprehend on some level too,” says Scholz. “Robots that comprehend: ‘Here are objects in an environment… What does it mean?”
To this end, the visionaries at the Australian Centre for Robotic Vision may team up via the D-CRC with an Australian centre of excellence in artificial intelligence. The agility afforded by working directly with Defence on these robotics projects will enable accelerated development, says Scholz, because rudimentary prototypes can quickly get input to make them functional. “It’s the only way I know of to develop technology that’s never been built before and where we don’t know how to do it — we just have a sense that it’s the right way to go,” he says.
Redefining the equation
D-CRC projects must be ahead of the curve to give Australia’s military the edge. With a relatively small defence force, a major investment in navy vessels and aircraft, and a large country to defend, the Department of Defence recognises the potential of Trusted Autonomous Systems to extend and augment its reach. “We talk in terms of ‘force multipliers’,” says Scholz.
Taking risks on projects that don’t just push the envelope, but “blow it apart”, are part of the centre’s remit, he says. Examples that might be considered include ‘tag-teaming’: inexpensive autonomous robotics underwater systems that can swarm the ocean floor in advance of multi-million-dollar Royal Australian Navy vessels, clearing a path through minefields or scouting out the bathymetry of the seafloor.
“We want autonomous systems that will survive in a warfare environment,” he says, and he’s not just talking about physicality. New kinds of platforms such as social and self-sensing are important as well, to enable effective interaction with humans.
This first CRC for Defence will operate for seven years, with options to extend if promising robotics projects need further development. The Brisbane headquarters are expected to be established in mid 2018.