Email this Article Email   

CHIPS Articles: Naval History and the Future Collide with Algorithms Built with Tabletop War Game Data

Naval History and the Future Collide with Algorithms Built with Tabletop War Game Data
By Elliot Carter, Naval Surface Warfare Center Dahlgren Division Public Affairs - October 30, 2020
DAHLGREN, Va. (NNS) -- In 1887, students at the Naval War College were astonished to find their teachers crawling over a game board on hands and knees, utterly transfixed by a collection of inch-long toy dreadnaughts. The concept of war games originated from the European land armies, but it was first applied to naval warfare here in the United States with the support of leaders who saw a real-world military advantage in Lilliputian-scale sea battles. “The fate of nations does depend on the movements of these toy ships,” one observer remarked in those early years.

An analogous miniature-themed exercise unfolded this year at Naval Surface Warfare Center Dahlgren Division (NSWCDD), where participants in the Sly Fox Mission 27 used a tabletop war game and custom-built computer algorithm to speed up parts of the force mix optimization analysis conducted at Dahlgren. As with the 19th century war games, Sly Fox organizers and participants believed that the exercise could have an outsized impact on the future of naval warfare.

Force mix is a data heavy field that compares the attributes of different Navy assets, and uses that information to ensure the Navy has a decisive advantage in future conflicts. The insights provided by force mix optimization help determine how the Navy prepares for war now, and informs shipbuilding investments that might take years or decades to come to fruition. Executing a force mix analysis can easily take a year to complete. However, analysis is known to start over around the halfway point if one of the underlying assumptions changes due to new intelligence or budget updates.

The objective for Sly Fox Mission 27 was to build a prototype algorithm that can handle the most labor-intensive aspects of analysis more rapidly and completely than human analysts. Developing a realistic prototype would need a large dataset. Rather than using actual technical information about the U.S. Navy ships, the group was able to use placeholder numbers better suited to the unclassified project. A popular commercial war game was selected as the best source for the placeholder numbers, due to the game manual’s “similar complexities to the U.S. Navy,” according to Josh Shiben, a product owner at NSWCDD. Bill Walsh, program director for Sly Fox added that the fantasy numbers “can be easily transferred with live Navy data” in future iterations of the prototype.

Building the project mimicking a popular game also had an undeniable appeal that the 19th century gamers would have recognized in their toy dreadnoughts. Namely, “some of us had a very heavy interest in tabletop war games,” said Kaela Gosdzinski, one of the Mission 27 participants. “There was a lot of background knowledge that we were able to bring in, and we already had the game manuals.”

The prototype that emerged from Sly Fox Mission 27 is a promising step toward real-world use. Dahlgren engineering and science participants built an algorithm that combines game units into miniature fleets, scores them, reshuffles the pieces and repeats the process. The program can run through this routine 20,000 times in three minutes. Each of the iterations increases the strength of the final recommendations for an ideal tabletop game fleet.

The designers describe this artificial intelligence (AI) capability broadly as a “genetic algorithm” because of the functional similarities to the process of natural selection. This also distinguishes the genetic algorithm from neural networks and many other types of AI tailored for different objectives.

Now that a proof of concept has been delivered to NSWCDD’s Mission Engineering (M) Department, the participants of Sly Fox Mission 27 hope that the project will continue to be refined into a process that can be used in real force mix optimization efforts. However, that is not expected to be an overnight process. “You still need to do a lot of analysis on the human end to check it,” Gosdzinski said. “The next version uses naval data and tests the strength of the conclusion.” Long term, she said using genetic algorithms on force mix analyses “has huge implications for the strength of the Navy.”

Sly Fox is a professional workforce development program at NSCWDD focused on junior Navy scientists and engineers. A competitive selection process selects seven individuals with the skills and experience best suited for the unique challenges that span hardware and software knowledge.

The team is responsible for shepherding the project along its entire systems engineering lifecycle, including concept, design, review, demo release, and a final product review. Program Director Bill Walsh describes the experience as [using] “five years of experience in six months.”

The participants in Sly Fox Mission 27 included Kaela Gosdzinski, E14 Physics; Sharon Adamavage, V22 Mathematics, applied statistics; Vijay Ahluwalia, H43 Electrical Engineering; Ibrahima Diallo, M31 Computer Science and Information Systems; Kevin France, V51 Computer Science; Noah Harlow, H53 Computer Science; and Travis Wright, R35 Mechanical Engineering.

Click for more information about Naval Surface Warfare Center Dahlgren Division (NSWCDD)

Participants in Mission 27 of the Sly Fox program at Naval Surface Warfare Center Dahlgren Division assemble for a team photograph. Pictured in the back row, from left to right: Kevin France, Ibrahima Diallo, Travis Wright, Sharon Adamavage. Front row: Vijay Ahluwalia, Kaela Gosdzinski, Noah Harlow. Photo provided by the Mission 27 team.
Related CHIPS Articles
Related DON CIO News
Related DON CIO Policy

CHIPS is an official U.S. Navy website sponsored by the Department of the Navy (DON) Chief Information Officer, the Department of Defense Enterprise Software Initiative (ESI) and the DON's ESI Software Product Manager Team at Space and Naval Warfare Systems Center Pacific.

Online ISSN 2154-1779; Print ISSN 1047-9988
Hyperlink Disclaimer