A U.S. Navy team won an inter-service racing competition hosted by Amazon Web Services and aimed to teach participants the basics of machine learning.
Naval Information Warfare Center Atlantic-led Team DeepBlueSea received the DeepRacer golden trophy after its victory against six finalists from the Navy and seven others from the Army in building the fastest ML-based autonomous race car, Dave Bader, lead business development manager at AWS, wrote in a post Tuesday.
Race participants used the Amazon SageMaker service to develop reinforcement learning models that were later applied to drive 1/18th scale vehicles during multiple timed races on a physical track.
The RL technique worked to assimilate each car with the racing environment through a reward system in which desired actions such as staying in a vehicle lane earn an incentive.
“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 Rob Keiser, a NIWC Atlantic senior scientific technical manager and DeepBlueSea leader.
Sai Liu, a software development lead at Army Communications-Electronics Command, noted that the competition offered basic information on ML that would come in handy should the service pursue the technology in the future.
DeepRacer is powered by the AWS RoboMaker cloud robotics service and designed to provide users hands-on ML learning experience that involves 3D racing simulation technology.
