Enable breadcrumbs token at /includes/pageheader.html.twig

Using Statistics to Conduct More Efficient Searches

Employing modern statistical inference tools can provide the Navy a bridge to improved searching and tracking.

Metron uses statistical methods to create inference tools for the Navy.  Metron Scientific Solutions

Metron uses statistical methods to create inference tools for the Navy. Metron Scientific Solutions

Employing modern statistical inference tools can provide the U.S. Navy a bridge to improved searching and tracking for people, planes and ships across land and sea. The ability of the tools to incorporate even spotty intelligence data to help locate individuals elevates the Navy’s command and control posture and ultimately aids in protecting the United States from security threats.

For help in this area, the Office of Naval Research (ONR) has turned to Reston, Virginia-based Metron Scientific Solutions as well as other contractors. In the undersea warfare community for 30 years, Metron has developed a search methodology for finding missing ships, aircraft or people at sea—as well as tools to track people of interest using intelligence data.

Applying its methodology, the company has helped in several high-profile searches. It assisted the French government in finding the wreckage of Air France Flight 447 in 2009 in the southern Atlantic Ocean and in the international search for Malaysia Airlines Flight 370 in 2014 in the Indian Ocean. Metron also helped build part of the U.S. Coast Guard’s Search and Rescue Optimal Planning System (SAROPS): the section that provides probability maps for locations of distressed ships or people, and plans effective searches. The Coast Guard uses SAROPS, in operation for the last 10 years, every day to plan searches, says Metron Chief Scientist Lawrence Stone. The company also is performing research and development for a probability-driven system that would help locate downed pilots behind enemy lines.

The key to these efforts is the use of a so-called Bayesian approach. “For us, the Bible is written in Bayesian,” Stone says. The approach mathematically identifies the probability of an event occurring based on prior observations and other related information. It evolved from 18th-century statistician Thomas Bayes’ theory that updates the probability of an event as more observations come in. The U.S. Navy first employed Bayesian search methods in World War II to try to stave off the threat of German submarines.

Metron’s latest work for the Navy and the ONR, the development of the Bayesian inference engine known as Integrated Long-term Recursive Tracker (I-LRT), harnesses intelligence data to help track or find people of interest. The I-LRT tool makes use of large amounts of information, including data that is spotty and has gaps or errors, says J. Van Gurley, Metron’s chief operating officer and a 26-year naval veteran. “This framework allows us to take into account hard information, such as a location report, even with some errors with it, and use it side by side with soft information, such as an intelligence assessment,” says the former submarine warfare office and naval oceanography specialist.

Stone, an expert in Bayesian multiple target tracking who is known as the godfather of the I-LRT, explains the methodology. “With the Bayesian approach, you start off with a prior distribution on the parameter you are trying to estimate and use observations to estimate a posterior distribution,” he says. “In this situation, the prior distribution corresponds to the stochastic, or probabilistic, motion model for the target you are trying to track.

“So you put all the prior knowledge of the target into the distribution with all the uncertainties too,” he continues. “And as you get observations, you get measurements and position estimates, and then you use those to revise your distribution of a target’s state at a present time, the time of measurement, and use it to project the target’s position at times in the future.”

Stone confirms that one of the tool’s strengths is its use of pattern of movement information. “It helps get you through the thin spots where you are not getting very many observations,” he says. “Observations can be kind of sparse, and you have to have some way of projecting ahead until the next observation comes in.”

Gurley adds that the system incorporates a set of hypotheses about what a particular red asset may be doing in the future. “It could be many different hypotheses,” he says. “It might be that they are going north, south, east or west, or they are patrolling or transiting, whatever they might be doing. Then we encode all that mathematically, with a set of prior observations, each with its own motion model and distributions as time progresses. As the system moves into operation, as contact reports come in, we can compare and contrast each of those pieces of hard information with the different hypotheses and the set of the things we thought the target might do. We can reason on that to say which of the hard pieces of information we should believe and which should we discount as well as which of the initial motion and mission hypotheses we should now believe more strongly than we did initially.”

The Metron executives offer that the method is an improvement on the standard fleet tactic of the farthest on circle. That tactic, still used today, results in a search circle “that, after not too many hours, gets so large that it is basically impossible to search,” Gurley says. The I-LRT tool provides an area much smaller than the farthest on circle.

“Instead, we have something that is based on where we saw them last, how we’ve seen them operating and what the intelligence is telling us to expect in the future,” he notes. “It merges all of that into a single probability heat map that says where the target most likely will be. The heat map continues to evolve as more observations come in, and it is always smaller than that farthest on circle.”

The ONR has shown the I-LRT to several antisubmarine warfare groups, which have used live operational data in the system, Gurley says. Metron’s initial feedback indicates that the tool has been favorably received. “It is still a developmental project, so anytime you put a new tool or capability in front of operators, you learn something from that experience, and that gives us information that we can go then refine, extend and improve,” Gurley says. Throughout the rest of the fiscal year, Metron has some specific design goals and tasks from the ONR.

“There is a lot of interest in the capability, and we have a lot of pieces going on, including to extend the capability,” Gurley shares.