ADEX: AI key to truly autonomous UAVs

Artificial intelligence (AI) will be a crucial enabler for UAVs to be truly effective in contested airspace, says a senior South Korean researcher.

Given the challenges involved with developing AI, however, it will be several years before UAVs that can operate truly autonomously will arrive, says Kang Wanguu, director of the Unmanned Vehicle Advanced Research Centre at the Korea Aerospace Research Institute (KARI).

“If a UAV is operating over enemy territory and loses its connection with the operator, it is no more than a two-year-old child that cannot find its way home,” says Kang.

It is therefore essential to move from a situation where there is a “man-in-the-loop” to one where there is a “man-on-the-loop.” AI is key to achieving this, but there is still some way to go.

Kang made the remarks at a conference in Seoul prior to the biennial Seoul ADEX defence show.

He classifies AI into three realms of increasing sophistication: narrow AI, general AI, and super-intelligent AI.

Narrow AI allows a UAV to perform specific tasks, such as landing and taking off automatically, locating lost hikers, identifying specific targets, and other similar objectives.

General AI allows for abstract thinking, and Kang reckons this could be decades away. Super-intelligent AI, where a machine is intellectually superior to a human brain is, he feels, the stuff of imagination.

“Unmanned fighters capable of aerial warfare are expected to be very useful weapons systems because they are capable of high G maneuvers, and save weight by eliminating the need for equipment such as ejection seats, life support systems, and control interfaces,” he says.

Wanguu adds that while such platforms offer significant promise, they will require a very robust wireless network that allows real-time transmission of imagery at high bit rates.

Korean Air’s Aerospace Devision (KAL-ASD) has done some preliminary work on a UCAV, but the project remains firmly on the drawing board.

Kang adds that another key trend in the UAV space is the diminishing size of sensors. This allows UAV platforms to be smaller, and by extension harder for adversaries to spot either visually or through other detection means, namely radar. Kang also sees great potential for swarming UAVs, whereby dozens, or hundreds, of small UAVs work together as a network to achieve missions.

“In a swarm, even if the adversary destroys a few units, it will not have a material impact,” he says.

While UAV technologies offer greater benefits, he warns that adversaries also benefit. When officials examined the hard drive of a small UAV believed to be of North Korean origin that crashed in South Korea, it was found to have taken hundreds of pictures of a US airbase.