Seven undergraduate students from the Indiana University Luddy School of Informatics, Computing and Engineering in Bloomington spent several days wrapping up their spring semester coursework from an unusual location: a small cluster of tents on a lakeshore in northeastern Indiana.
The students — along with graduate student and faculty mentors — were participating in the inaugural Artificial Intelligence Maritime Maneuver Indiana Collegiate Challenge, an invitation-only event hosted by Trine University that tasks college students with designing, programming and launching an autonomous, AI-powered watercraft. Supported by the Office of Naval Research, the challenge provides students with a low-cost, easy-to-build, low-profile vessel with the capacity to be made fully autonomous.
Participants in the competition are required to adapt the vessel — a 14-foot unmanned skiff — to operate independently using artificial intelligence. The IU team developed and installed its technology in a hanger-like space in the Multidisciplinary Engineering and Science Hall on the IU Bloomington campus. They then spent three days at Pokagon State Park, where the event took place, splitting time between a lodge and lakeside campsite.
“This was really a ‘choose-your-own-adventure’ type challenge,” said David Crandall, Luddy Professor of Computer Science and director of the Luddy Artificial Intelligence Center. “We essentially received a ‘boat kit’ with a pile of equipment — sensors, cameras, radar, computers, motors, a steering mechanism — and the students had to figure out how to put everything together and make it work; it’s not like it came with an instruction manual.”
IU team members were undergraduate students Adebowale Adelekan, Mathew Arbuckle, Gautam Hari, Jackson Miller, Sparsh Nair, Tri Huu Minh Nguyen and Sotaro Kaneda; graduate student mentors Paul Coen and Michael Siler; and faculty mentors Crandall, Memo Dalkilic and Lantao Liu. All seven undergraduates were also members of the Luddy AI Club, a student-interest group focused on AI theory and practice.
The other five teams participating in the competition came from Purdue University, Purdue University Fort Wayne, Rose-Hulman Institute of Technology, Trine University and University of Notre Dame. IU placed third in the technical design report category.
“I was proud of the students who participated in the event,” said Coen, the primary mentor who spent two months coordinating and assisting the undergraduate students with both hardware assembly and software development. “Everyone really threw themselves into the challenges, and there was a great spirit of camaraderie.”
“Paul’s commitment and guidance were invaluable to the project’s success,” Crandall said. “I can’t think of another event where so many students and faculty from so many other universities get to spend real time together learning and collaborating.
“This wasn’t a cut-throat competition; everyone was encouraged to share tools and insights,”
Starting in February, the IU team had about eight weeks to work through the competition’s hardware and software challenges. The final sprint in the competition’s last three days was the team’s first opportunity to put the boat in the water. It was also a physical challenge in some ways, with team members regularly hauling the boat in and out of the water, testing the motor, even loading 1,000 pounds of ballast into the hull.
“April in Indiana isn’t exactly the Caribbean in terms of water temperatures,” Crandall joked.
A rising junior in computer science, Nguyen said the project provided participants the chance to implement AI systems such as computer vision — or the ability to interpret information from the cameras accurately — as well as overcome numerous hardware/software interaction challenges, such as ensuring the boat would react appropriately based upon input from its sensors.
“This project really helped me gain new skills and perspectives on the field of computer vision,” he said. “We were able to learn a lot from our mentors and I gained a lot of knowledge on topics that I would like to continue to pursue.”
The IU team was one of the few that got its computer vision system up and running properly, Crandall added. Some teams’ boats never even got out of the dock, he said.
“There were some problems that were simply mission critical — that if they didn’t happen, we wouldn’t have a boat in the water,” said Gautam, also a rising junior in computer science. “So we simply got it done, even if it was way outside our current knowledge areas. It really helped us practice learning quickly.”
Among the competition’s tasks were navigating in and out of the dock, traversing a “slalom course” made up of floating buoys, collecting rings along a path in the water, detecting buoys of specific colors, deploying sensors and sensor communications, and a simulated “search and rescue” operation.
Each team also delivered a presentation to a panel of experts from organizations such as NSWC Crane and the Office of Naval Research.
Although the competition was first and foremost a learning exercise, Crandall said that AI and machine learning will play a critical role in real-life autonomous vehicles. Applications of the technology may one day include unmanned search-and-rescue operations or the ability to map treacherous waters without endangering human workers. At IU, the Vehicle Autonomy and Intelligence Lab, led by Liu, is among the groups actively working to make this future a reality.
“Learning to navigating through an unconstrained environment like a lake — where you can’t control the wind or the waves — is a really important AI research challenge,” Crandall said.
A lake is also the perfect place to practice since it’s relatively enclosed, compared to the open road with a self-driving car or the open skies with a drone, he said.
“I don’t know if anyone quite realized the scale of the challenge until we started working on it,” Nguyen said. “We took a project that could have taken months and compressed it down to a few weeks; it was a true learning experience.”