ABERDEEN PROVING GROUND, Md. -- A pair of teams representing the U.S. Army Communications-Electronics Command’s Software Engineering Center placed in the inaugural AWS DeepRacer Army-Navy Competition, a two-part event designed to teach machine learning in a dynamic and competitive environment.
The competition taught elements of reinforcement learning, an advanced machine learning technique that is ideal for training autonomous vehicles that compete in the AWS DeepRacer events.
Using Amazon Web Services software, 23 teams were tasked with building and training car models to compete against Navy and Army racers. A semi-final qualifier race was held March 11, in which the top seven Army and Navy teams each advanced to the finals.
For the final race, the models the teams developed were downloaded into racecars and raced along the physical AWS DeepRacer – PA Prime Loop track. The race was streamed live so teams could watch their cars in action.
The “Green Machine from Aberdeen,” a team made up of SEC Technical Services Directorate members Jerry Acord, Andrew Graham, Bob Clingan and Tara Diaman, placed second within the Army teams and seventh overall in the finale, completing the course in 15.768 seconds.
Acord, a project lead in CECOM SEC’s infrastructure service division’s software engineering environment (SE2) team, said the event was useful in that it helped broaden the team’s technical expertise and it was a good team building experience. He said he could come across customers who need support in machine learning.
“It would behoove us to expand our horizons a little bit about (machine learning) in advance,” he said.
Clingan, a teammate of Acord’s on SE2, said as the team optimized the car, it used machine learning functions to reward the car for how it behaved around the track. The car’s thinking can change, depending on the style and shape of the track it’s on, he said.
He was surprised to see how the physical car performed during the finals, compared to how the car functioned in the virtual setting.
“It has cameras and takes pictures [of where it’s going],” he said. “Little things can throw it off.”
A second CECOM SEC team, “Command and No Controls,” featuring SEC Command, Control and Communications Tactical Directorate members Sai Liu, Morgan Hansen, Matthew Holtsinger and Seiichi Sugawara, placed sixth within the Army and 12th overall, with a time of 28.985 seconds.
Liu, a software development lead at CECOM SEC C3T’s mission command division, said machine learning was something his entire team was curious to learn more about. While none of the systems he helps sustain use machine learning at the moment, what he learned could come in handy, moving forward.
“I think it’s good to learn about the new technologies, so in the future, if there’s anything that can allow us to apply our knowledge, when we have a little background to start,” he said.
Kim Bowers, workforce development strategist for CECOM SEC, said the competition was a way for employees to enhance their professional development.
“It was a great opportunity for our employees to learn more about Amazon Web Services and their machine learning capabilities,” she said.
Despite the Army teams’ best efforts, the grand prize winner, finishing in 7.112 seconds, was a Navy-based team, DeepBlueSea.
Social Sharing