Citadel cadets crossed the finish line in victory as they outraced West Point and Naval Academy cadets during the Amazon Web Services DeepRacer Tournament by using the power of Artificial Intelligence, or AI.
Through support from Amazon Web Services, or AWS, The Citadel hosted the first 2023 Senior Military College and Service Academy Warrior Week DeepRacer Tournament. Prior to the event, cadets from The Citadel, the U.S. Military Academy at West Point and the U.S. Naval Academy trained AI models to drive AWS DeepRacer cars autonomously, meaning on their own, using a video camera to “see” the track. The winning team was determined by which program guided the car to complete a lap the fastest.
The Citadel’s team consisted of Cadets Frederick Vogel, Blakeley Odom and Brian Bradrick, as well as their faculty coach Pooya Niksiar, Ph.D.
“Being a Mechanical Engineering major, I was apprehensive about going into an AI-based competition for the first time. It was all a relatively new concept for us and there was a lot of success and failure along the way, and we learned quite a few lessons that apply in real life as well as in competition. When we finally saw how well our machine performed in the real world, we were flooded with relief that all of our work had paid off in the end, and I can’t wait to help prepare the competing team next year to push the limits of what those machines are capable of,” said Odom.
Just how did these cadets train the car? Practice, and lots of it. Starting in the summer, they began training the car virtually on a cloud-based simulator and then downloaded the AI models to the DeepRacer car, which were used to drive the car on the track. The car navigated the track based on what it learned from experience and trial and error in the simulator until it perfected what it was doing. The algorithm they used was based on deep reinforcement learning. The purpose of this competition, including the training leading up to it, was to use DeepRacer as an AI teaching tool. The car drove around the track based on a model that it learned by training on deep neural networks, which took place well before the race began.
Andrew Williams, Ph.D., dean of the School of Engineering explained reinforcement learning as similar to training a dog.
“You give a reward for good behavior, and you don’t give an award for bad behavior. So based on how the car sees images and does an action based on those images, the reinforcement algorithm either rewards good actions or punishes the car’s bad actions over the long term, like running off the road. These types of algorithms can be used in financial services, chemical processes and we are demonstrating how it is being used for autonomous driving.”
AWS DeepRacer offers participants an engaging and fun way to begin using reinforcement learning, or RL. Compared to other machine learning techniques, RL is an advanced machine learning technique that takes a different approach to training models. RL can learn complex behaviors without requiring any labeled training data and can make short term decisions while optimizing for a longer term goal.
“This was the first year we were participating in the AWS DeepRacer competition with other senior military colleges and service academies, and we decided to host it as well, so a lot of work was put into ensuring it would be a great event. I am proud of The Citadel team for their hard work and dedication — the Naval Academy and West Point had participated in this competition before and had some prior experience and knowledge, while we started from scratch in the summer of 2023,” said Niksiar. “Our team did a great job not only in getting ready in a short amount of time, but they also took the first and second place in the competition. DeepRacer is a great way for our cadets and students to get involved with Artificial Intelligence, specifically reinforcement learning. In the near future, AI is going to be an indispensable part of research and industry and it is critical for our students to acquire the skillset in this area.”
The School of Engineering’s Center for AI, Algorithmic Integrity and Autonomy Innovation, or AI3, aims to bring AI teaching tools to Citadel cadets, students, the community and the state. In addition to AI3, the School of Engineering partnered with AWS Machine Learning University’s AI Educator Enablement Program to put Citadel faculty, including Niksiar, through Advanced AI and Machine Learning boot camps, ensuring that the faculty can teach machine learning.
“I am very proud of Dr. Niksiar and his team for taking first place and second place against their competitors. The Citadel thanks the U.S. Military Academy and the U.S. Naval Academy for their strong competition,” said Williams. “We are happy that The Citadel School of Engineering can provide a world-class engineering education here in the Lowcountry and in the state of South Carolina to serve our state and globe with principled leadership and engineering innovation.”
The Citadel has been in the Top 25 Undergraduate Engineering programs since 2012 and is one of the first five engineering programs in the country, providing a superior experience and education to each cadet and student.