Scientists at the University of Cincinnati successfully trained drones to autonomously land on moving targets. Researchers say “fuzzy logic” was the key to their success.
Landing a drone on a moving target requires precision.
“It has to land within a designated area with a small margin of error,” Manish Kumar, associate professor of mechanical engineering at the University of Cincinnati, said in a news release. “Landing a drone on a moving platform is a very difficult problem scientifically and from an engineering perspective.”
Such precision is difficult, so researchers looked to conquer the feat with a looser approach. “Fuzzy logic” describes a foil to exactness; it describes the use of generalities and educated guesses. Whereas exactitude leaves no margin for error, fuzzy logic allows for nuance and shades of accuracy.
“In linguistic terms, we say large, medium and small rather than defining exact sets,” Kumar said. “We want to translate this kind of fuzzy reasoning used in humans to control systems.”
When incorporated into an algorithm for drone navigation, fuzzy logic become “genetic logic.” It allows the system to constantly evolve and throw out the least useful solutions. Researchers used their new software to land quadcopters on landing pads mounted on moving robots. They named their fuzzy logic-based system ALPHA.
“Compared to other state-of-the-art techniques of adaptive thinking and deep learning, our approach appears to possess several advantages,” said Kelly Cohen, aerospace engineering professor. “Genetic fuzzy is scalable, adaptable and very robust.”
Researchers say their work could be used to land drones on aircraft carriers.
“This landing project is a real-world problem,” said researcher Nicklas Stockton. “A delivery vehicle could have a companion drone make deliveries and land itself.”