Army to use artificial intelligence to predict which vehicles will break down
By AARON GREGG | The Washington Post | Published: June 26, 2018
The U.S. Army is turning to a little-known tech start-up to automate the job of an equipment technician: It will employ artificial intelligence to flag failing vehicle parts before they break down in combat.
The company, Chicago-based Uptake Technologies, has finalized a $1 million contract agreement with the U.S. Army, under which the company's technology will be tested on a few dozen vehicles before they decide whether to scale it up for broader use.
Uptake's artificial intelligence will be applied to deployed Bradley M2A3 combat vehicles, an armored infantry transport vehicle manufactured by BAE Systems, a British defense contractor with a U.S. office in Arlington, Virginia.
"We're looking to see if we can leverage some of Uptake's machine learning algorithms to spot equipment failures before they happen," said Lt. Col. Chris Conley, the Army product manager for the Bradley fleet. "If this pans out and can provide some real capability, the Army could look to expand this to the entire Bradley fleet as well as other combat vehicle fleets."
Military leaders have been outspoken about the need to apply advanced artificial intelligence to the military's operations, but efforts to forge partnerships with Silicon Valley tech firms have been fraught with difficulties. An employee revolt recently prompted Google to back out of a project to rapidly process drone footage, for example.
Uptake will instead apply its AI to the "readiness" side of the equation. The services have come under criticism in recent years for spending profligately on expensive new hardware such as the F-35 Joint Strike Fighter while older, more practical vehicles fall into disrepair.
Uptake's trial run was set up through a sole-source contract arrangement coordinated by the Defense Innovation Unit Experimental, known as DIUx, an agency created under the Obama administration to forge ties between tech companies and military agencies. The contract amount of $1 million could lead to more business for Uptake if the trial goes well, a person familiar with the deal said.
Uptake has made its business analyzing the masses of complex signals that come off industrial equipment, crunching the data to keep tabs on equipment components and give factory managers better insight into the state of their equipment. In some cases, the company places tiny sensors on industrial equipment to collect data, but with the Army's vehicles the company will simply be analyzing the signals that are already produced by industrial equipment.
"Just like humans have been putting their statuses on Facebook and Twitter, these machines have been putting out their statuses for decades and nobody's been listening. Only recently do we have the technology to understand that," said Ganesh Bell, president of Uptake Technologies.
Officials are taking a cautious approach at first. In the three years since its founding, Uptake has accumulated an impressive docket of industrial customers, including Boeing, Caterpillar and an undisclosed national trucking firm. But it is yet to be tested on military vehicles, leaving the Army in uncharted territory.
"I'm not convinced that this will be successful, but I'm really excited about the potential of it," said Conley, the Army product manager. "We're doing a pilot test to verify their claims before we do anything at scale."
The company has argued that its wealth of experience and data from working with similar vehicles will give the company a strong starting point as it begins work with the Army.
The start-up benefits from the direct involvement of retired U.S. Navy Adm. Mike Mullen, who served as chairman of the Joint Chiefs of Staff under Presidents George W. Bush and Barack Obama. Mullen said he owns shares in the company and is consulting on military matters as a strategic adviser.
In an interview last week, Mullen said Uptake's process could drastically improve upon the military's current methods.
"What I've seen on the component side is you almost wait for failure and then figure it out," Mullen said. Uptake's analytics "will give you much better information on what your maintenance system should be. It allows you to be precisely predictive on when a part is going to fail, when a component is going to fail, when the whole system is going to fail."
Mullen said it could one day grow beyond the Army's basic pilot phase: "Based on the results I've seen there is a huge potential here for better outcomes and a lot less expense, which is what anybody in the military is focused on."