A 3D Globe for Operational Training

With the aid of machine learning, One World Terrain strives to enhance military training


The slogan “train like you fight, fight like you train” has long been a military mantra. But any soldier who has walked alongside model buildings made of cinder blocks at Fort Polk and been told to imagine themselves on an operation, only to find themselves in Syria a month later, can relate to the difficulty of mentally transitioning from training to live fire.

Full realization of the mantra requires accurate geospatial data and digital support to create a more realistic environment for training, then applying said data to conduct the mission.

The quest to create a 3D, high-resolution map intends to take some of that imagination out of the equation.

“One World Terrain” (OWT), a research effort underway as part of the U.S. Army’s Synthetic Training Environment, would “expand the understanding of geospatial data at high resolution to the whole planet,” said Barry Tilton, who moderated Sunday’s GEOINT Foreword panel discussion titled, “Working with GEOINT at Scale: Leveraging Machine Learning.”

With the map as a center point, the data it generates, along with the proliferation of other data sources, can create a training model that reduces surprises in battle.

The operative word here is “scale.” Or, as panelist Isaac Zaworski of global 3D data and analytics provider Vricon put it, “human scale.” “The scale is part of the bigness of the problem,” he said.

The small, discrete databases often used by artists to create the limited fidelity used in today’s training must give way to big data that can drive a new, geo-specific reality for the training of the future, according to Zaworski.

“We’re focused on collecting real data about the real world.” he said.

The data is overwhelming and will require help to both develop and deploy it. The goal is to reduce the weeks or months and millions of dollars required to accurately model a city for training purposes to hours at a lesser cost.

Needed are machine learning algorithms that can mesh elements of segmented data—about buildings, forestry, ground conditions, etc.—that would impact the mission, then turn the result into a realistic fighting environment. Those algorithms can guide the entire process, from data collection to creating the simulated training environment.

The goal is to move from the semi-automated procedure used today to one that is fully automated. Doing so will “improve the accuracy … without the constraints of the human workforce that is available … and can be easily scaled up,” said Meida Chen, a graduate research assistant at the University of Southern California, which has done extensive research as part of OWT.

But the path toward Army adoption of OWT is a long one, according to Joshua Delmonico, enterprise support branch chief at the Army Geospatial Center.

“I’m in the enviable position of seeing all of the technology and things that are possible,” Delmonico said. “And part of my mind says, ‘Yeah, but …’”

The “but” includes the ability to integrate with current systems, many with out-of-date components.

“With the One World Terrain, we have to figure out how to take all of that data, which is generally 2D … and feed it into a variety of systems,” Delmonico said. “And one of the big challenges is how to take this fundamentally 2D-designed system and 3D data and make those things work for that ‘train to fight’ paradigm.” 

Headline Image: Panelists from left to right: Barry Tilton, USGIF volunteer; Isaac Zaworski, Vice President, Vricon; Meida Chen, graduate research assistant, Computer Science & Civil Engineering, University of Southern California; Joshua Delmonico, Enterprise Support Branch Chief, U.S. Army Geospatial Center


Creating Shared Consciousness

Gen. Stanley McChrystal’s prescription for today’s complex world