Joshua Riojas clearly remembers the day he ran into a specific constructing at St. Mary’s College a number of years again. He was taking a self-guided campus tour, he mentioned, when he all of the sudden noticed an enormous, blue, steel construction with a phrase in its title that sparked curiosity: Unmanned.
It was the Unmanned Aerial Techniques Lab, a lab constructed by the college in 2021 to include drones and drone-related analysis into their Science, Engineering and Know-how curriculum.
Riojas, now a 22-year-old laptop engineering main at St. Mary’s, knew he needed to work on this lab. What he didn’t know is that solely a few years later he can be growing machine studying fashions — an AI subfield that seeks to coach algorithms to make predictions primarily based on knowledge and statistical strategies.
His analysis seeks to assist maintain unmanned aerial autos up within the air longer, in hopes these can someday be deployed for prolonged survey or search missions, like these taking place after the floods in Kerr County.
“I’m fairly positive you will have seen on social media, ‘Why wouldn’t they deploy drones for search and rescue missions?’” Riojas mentioned. “And the straightforward reply is that they’ll’t final lengthy within the air as a result of energy points.”
Engineers at St. Mary’s College helped safe a grant in October 2024 totaling greater than $340,000 from the U.S. Division of Protection to buy autonomous autos, together with flying drones and floor autos for this and different analysis initiatives.
“This grant allows us to do analysis on autonomous autos, AI and general safety,” mentioned Wenbin Luo, Ph.D., professor of laptop engineering at St. Mary’s. “As well as, we additionally wish to use it as a educating platform, so we will use what we study and go the information on to our college students.”
Whereas Luo and different professors and college students deal with analysis to maintain autonomous autos secure from cyberattacks, the query for Riojas grew to become: How do you assist maintain these machines up within the air longer?
“The one resolution to having a system that may keep aloft within the air for an extended time frame is a sailplane, as a result of they use the wind to remain aloft for an extended time frame,” Riojas mentioned.
That meant he had to determine easy methods to train the machines to learn air currents, search air patterns and use these patterns like a hen would, to remain in flight with out expending an excessive amount of power.
“On this case, Joshua is creating knowledge,” mentioned Ben Abbott, the engineering lecturer overseeing Riojas’ work within the lab. “He’s making an attempt to create a machine studying — on this age of AI that we’re in — that may change into as good because the hen, smarter than people.”
The U.S. Division of Protection affords grants to increased schooling establishments to help analysis that may be crucial to nationwide safety. In fiscal yr 2025, the company spent greater than $5 billion on grants and awarded greater than 2,000 new grants, in response to knowledge launched by USAspending.
On this case the grant was secured by St. Mary’s to assist help cybersecurity analysis, particularly easy methods to maintain unmanned autos secure, however having this tools affords a wide selection of different analysis prospects, Abbott and Luo mentioned.
In his analysis Riojas created knowledge by studying air patterns himself, such because the course of a present, the speed of it, and mixing this with how a drone would react to a selected present; for instance, the place within the explicit present would a drone fly in a extra steady approach? What knowledge factors may he collect from even probably the most minuscule motion attributable to air fluctuations or change in place?
To try this, he connected movement caption markers to a flying drone and requested it to observe a car on the bottom with a fan connected to it utilizing sensors. He then recorded the information it was receiving from the drone’s motion together with its actual location inside that rising air present.
“I simply discovered it very fascinating how correct the machine studying fashions that you just construct may be,” Riojas mentioned. “If that drone is on the very middle of that rising air, then it’ll inform you that it’s on the middle. But when it’s on the fringe of it, then it’ll inform you as a result of it’s going to be wanting on the roll, pitch and yaw.”
Along with his professor wanting attentively, Riojas managed solely the bottom robotic to maneuver across the flight space within the lab, blowing air because it roamed. A small drone then adopted its motion intently, searching for to remain close to the middle of the rising air because it gathered knowledge with its actions.
With this analysis Riojas and his professors hope to someday inform many different unmanned aerial autos easy methods to keep within the air longer and make a distinction in doubtlessly life-saving missions.
That is additionally one instance of why Abbott sees working with AI as a requirement in his programs. He desires to offer his college students hands-on expertise in order that they study to make use of AI as a software to unravel real-life drawback.
“Nicely, what we’re studying right here … is easy methods to use the AI,” Abbott mentioned. “I imply, Joshua, he doesn’t understand it but, however he can compete with anybody.”
