The military will increasingly rely on autonomous and unmanned systems in future conflicts, but autonomous aerial platforms are developing at a faster pace than their ground-based counterparts, said Deputy Secretary Bob Work on Wednesday.

Although commercial companies like Tesla Motors [TSLA] and Google [GOOG] are pouring millions of dollars into driverless cars, those vehicles do not meet military requirements for an autonomous vehicle capable of moving offroad in remote locations where precision navigation is more difficult, he said.

“In the commercial space, they generally are going after cars that would operate primarily on paved road,” he said during an event hosted by The Washington Post.  “Whereas the military, not only will we stay on roads, but when the roads become more dangerous, we’ll go offroad. That type of navigation is extremely difficult.”

Oshkosh developed its TerraMax technology for autonomous convoy operations. (Photo: Oshkosh).
Oshkosh developed its TerraMax technology for autonomous convoy operations. (Photo: Oshkosh).

“I would expect us to see unmanned wingmen in the air before we would see unmanned convoys on the ground,” he added, pointing to an Air Force concept called Loyal Wingman in which an unmanned QF-16 works together with an F-35, which Work called “a fifth generation battle network node.”

The military has also demonstrated the use of unmanned aerial vehicles to deliver cargo to soldiers in Afghanistan, he said, referencing the K-MAX helicopter operated by the Marine Corps starting in 2011. It returned to the United States in 2014 after flying thousands of resupply missions.

Work also expects to see further development of autonomous maritime systems, such as distributed undersea systems and unmanned boats, he said.

The Army continues to see autonomy as one of the most promising technologies coming out of the commercial sector because of its applicability for unmanned ground vehicles, Scott Davis, program executive officer for combat support and combat services support, said earlier this month at the Association of the U.S. Army Global Force Symposium and Exhibition in Huntsville, Ala. However, it will be a while before the service fields fully autonomous vehicles.

The service likely would first deploy vehicles with drive-by-wire active safety kits that help drivers avoid collisions and stay in their lane, creating an underpinning architecture for autonomous capability in the future, he said.

From there, it would field a leader-follower capability, in which a manned vehicle would lead a convoy of autonomous vehicles programed to follow the one in front, Davis said. Defense companies such as Oshkosh [OSK] are already developing and demonstrating that capability, which could be on Army vehicles sometime this decade.

A fully autonomous convoy capability could follow in the mid to late 2020s, he said.

On Wednesday, Work said the Defense Department’s conception of autonomy was simple.

“It is nothing more than delegating decision authority to an entity in…a battle network,” he said. “And a battle network is nothing more than a sensor grid and a command and control and communications and information grid and effects grid, and they’re all kind of stacked together. You’re finding out what’s going on in your environment, you make a decision on what effects you want to achieve, and your network can do that.”

In the next five to 10 years, the Pentagon will have a very limited, narrow use of autonomy and artificial intelligence capabilities. For example, “things like pushing a button and doing a parallel park with your car,” Work said. The department will not give machines the ability to make decisions that could cause loss of life, but it could give them some decision-making authority in areas like cyber or electronic warfare, where computers can move faster than human reaction time.

Another key technology is “deep learning machines” that can quickly crunch data from a host of sources and identify connections that human users can analyze. The department will need such capabilities to further its understanding of the Islamic State’s financial and organizational structures so the military can outmatch it, Work said.

“We are absolutely certain that the use of deep learning machines is going to allow us to have a better understanding of ISIS as a network and a better understanding on how we can target it precisely and lead to its defeat,” he said, using another name for the Islamic State.

A Silicon Valley company recently demonstrated its own analytic technology to Work and other department officials that compiled photos and video showing evidence of Russia firing a surface-to-air missile in 2014 to shoot down Malaysia Airlines Flight 17, using Twitter, Instagram and other sources where users posted information in real time.

The company was able to sort through millions of pictures available online that a human analyst would have taken longer to parse, he said.

“When you add deep learning machines and computers which are designed to help analysts understand the connections that they’re seeing, the analyst will become better, they will provide better input to their commanders, [and] the commanders will understand the effects that they want to achieve,” he added. “So we are very, very excited about the prospect of working these things into our fight.” 

Dan Parsons contributed to this story.