The two-part event, which centered on racing autonomous vehicles, took place March 11-25. The competition began with a workshop and community race and ended with the top 14 finalists – seven from the Army and seven from the Navy – competing for the top spot in the virtual challenge.
Of the 23 teams that participated in the competition, NIWC Atlantic supplied eight teams composed of 26 data scientists and artificial intelligence (AI) engineers to represent the Navy.
Six NIWC Atlantic teams placed in the top 10 and five of those teams claimed the top spots in the DeepRacer competition. The winners received virtual trophies as well as “bragging rights for clocking the fastest lap time,” according to the AWS DeepRacer website.
“NIWC Atlantic is committed to being a leader in data, analytics and AI workforce for the Department of the Navy,” said Senior Science and Technology Manager Robert Keisler, who is responsible for more than 250 personnel as leader of the Data Science and Analytics Engineering team at NIWC Atlantic. “One of our strategic investment areas is focused on autonomy, and reinforcement learning (RL) that drives the DeepRacer vehicles is a key component of that.”
The AWS DeepRacer website states that DeepRacer gives developers a simple way to learn RL, experiment with new RL algorithms and simulation-to-real domain transfer methods and experience RL in the real world.
“The key aspect of AI for DeepRacer is reinforcement learning. This is essentially how you would train a dog,” said Ryan Vinson, NIWC Atlantic data scientist and DeepRacer team lead. “When the robot does something correctly, you give it a reward. When the robot does something wrong, you either correct it, or give it a smaller reward. The robot wants to get as much of a reward as possible, and so, hopefully, learns to take the correct action.”
“The key to all of this – and what makes it AI – is that the robot learns all of this on its own. We don’t ever explicitly tell it to do something or not do something,” Vinson said.
Artificial intelligence is defined as the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making and translation between languages. According to NIWC Atlantic Data Scientist Lucas Overbey, AI is one of the warfare center’s eight technical focus areas, making it a major emphasis of research, development and engineering efforts.
“NIWC Atlantic is working to be at the forefront of data science, machine learning (ML) and AI science and technology within the DoD — including establishing the Analytics Research Center (ARC) in 2015, the Seaworthy Artificial Intelligence Lab (SAIL) in 2019, and hosting the DON Data Science and Analytics Workshop in 2018 and 2019,” said Overbey, who participated as a coach in the DeepRacer event.
Machine learning is the use and development of computer systems that are able to learn and adapt without following explicit instructions using algorithms and statistical models to analyze and draw inferences from patterns in data.
DeepRacer is an autonomous racecar, approximately 1/18th the size of a real car, that uses cameras to view the track and an RL model to control throttle and steering. It is the first autonomous scale car specifically developed to help developers get hands-on with RL, an advanced ML technique that is “ideal for training autonomous vehicles that compete in the AWS DeepRacer events,” according to their website.
Participants utilized Amazon SageMaker – a service used to build, train and deploy ML models for use cases – to build and train models for the event. During the finale, racers watched via livestream as their models, which were downloaded to AWS DeepRacer autonomous vehicles, physically raced at the AWS DeepRacer - PA Prime Loop.
NIWC Atlantic currently has three physical cars in their possession, which will be used to continue investments in RL and workforce development in the coming months.
“This competition gave individuals of all skill levels a chance to get their hands on real-world devices and use their previously learned skills to create AI to drive a car around a racetrack,” said Cameron Arnold, NIWC Atlantic scientist and DeepRacer team lead.
According to Arnold, more than half of the NIWC Atlantic participants had no prior experience with the type of AI and ML utilized in the DeepRacer challenge. “The hands-on experience of the DeepRacer competition will bolster the command’s ability to create models that help with autonomous decision-making in the future,” he added.
Plans for AI and ML continue to expand at NIWC Atlantic, to include exploring how RL techniques can be applied to a wide range of naval problems such as autonomous navigation; command and control; scheduling and logistics; automation and condition-based maintenance via the Naval Innovative Science and Engineering (NISE) program.
“This event was a great learning experience and one of many ways we will ensure the Navy leverages data and AI to achieve decision superiority in warfighting and business,” said Keisler, who was named Grand Prize Winner for the record-breaking time of his autonomous vehicle, DeepBlueSea.
For more information about DeepRacer, see https://aws.amazon.com/deepracer/league/.
As a part of Naval Information Warfare Systems Command, NIWC Atlantic provides systems engineering and acquisition to deliver information warfare capabilities to the naval, joint and national warfighter through the acquisition, development, integration, production, test, deployment, and sustainment of interoperable command, control, communications, computer, intelligence, surveillance, and reconnaissance, cyber and information technology capabilities.