YUMA PROVING GROUND, Ariz. — Army senior leaders say that successful deterrence against near-peer adversaries with the ability to conduct large-scale combat operations will require long-range precision fires, autonomous capability, and leveraging new technologies across all military branches.
The Army wants to reduce sensor to shooter timelines, react to threats faster, and combine all of the systems and effects available at their disposal.
It is a tall and complex order, and U.S. Army Yuma Proving Ground is at the forefront of conducting developmental testing of the equipment that may be used by the future force.
In conjunction with the Army Test and Evaluation Command it is subordinate to, YPG is developing the local architecture and establishing data governance in advance of more practical case uses for artificial intelligence in support of the post’s test mission.
“The big thing with AI is being able to introduce time efficiency, to use data and build models where it will learn from information it is provided and do some process that a human would do, but more efficiently,” said Richard Hernandez, YPG’s Chief Data and Analytics Officer. “It’s really about identifying the pain points in your process — when you understand your data and your process and get to the point where you are doing some kind of predictive analysis, that is where AI and machine learning really helps you out.”
Data is in a very real sense YPG’s chief product, and the post’s Data Processing Branch is the nerve center for reducing the enormous amount of data collected during tests on the post’s vast ranges: optical data, tracking radar data, ballistics data.
“We also get the occasional weird thing that only comes every five or ten years, but it isn’t like a private industry business where we say we aren’t going to provide that to a customer anymore because there isn’t enough demand for it,” said Ashley Thompson, chief of the branch. “If the Army needs that thing, we have to make sure we have the expertise and the tools to process the data. That’s hard when people retire and take that knowledge out the door: we have to be sure that even if we automate it, we understand how the automation works.”
The benefit that YPG has over other organizations or businesses is troves of historical data from decades of test events. This data is extremely valuable for training AI models to automate or expedite data reduction and analysis. These models are already resulting in increased accuracy and significant time savings in achieving data products that support system performance assessments. A recent successful example involved developing a workable algorithm to help facilitate the acoustic trilateration of air to surface missiles and other helicopter rounds collected from arrays of microphones and hydrophones on the post’s highly instrumented ranges.
“When there are six to eight submunitions going off at different times, but close together, it can be really hard to triangulate the impact of all of them,” said Thompson.
With the help of longtime YPG analysts, a developer created a program that reduced processing time first from months to days, and finally to seconds.
“We should be automating the rote, redundant tasks and let the analysts do the more interesting analytical work,” said Thompson. “Some of it isn’t AI, it is just simple automation.”
Since munitions and weapons testing has long been the post’s most significant workload, the branch wants to autonomously process things like the test data generated from lot acceptance testing of mortar rounds.
“We want to facilitate that real-time data, so we are working on a statistical model and building up infrastructure to get them that information immediately in the field,” said Thompson. “If there’s an issue with a particular set of data looking questionable, we would post-process it here.”
Social Sharing