Dung beetles drive AI vision
Print
Engineers have taken inspiration from the dung beetle's unique navigation abilities to develop a groundbreaking AI-based navigation system.
The research by a team at the University of South Australia could revolutionise the guidance systems of drones, robots, and satellites.
The dung beetle, an insect that has been around for 130 million years, is the first known species to navigate by using the Milky Way.
This discovery, made by Swedish researchers in 2013, revealed that dung beetles rely on the light from the Milky Way to roll their dung balls in a straight line, ensuring they move away from competitors.
Building on this discovery, an Australian engineering team led by Professor Javaan Chahl has used the beetle's celestial navigation technique to create an AI sensor capable of accurately determining the orientation of the Milky Way even in low-light conditions.
This sensor is designed to overcome the challenges of motion blur that typically affects the visibility of individual stars during movement.
“Nocturnal dung beetles move their head and body extensively when rolling balls of manure across a field, needing a fixed orientation point in the night sky to help them steer in a straight line,” Professor Chahl says.
“Their tiny compound eyes make it difficult to distinguish individual stars, particularly while in motion, whereas the Milky Way is highly visible.”
In their experiments, the team mounted a camera on the roof of a vehicle to capture images of the Milky Way both while stationary and in motion.
These images were then used to develop a computer vision system that consistently measures the Milky Way's orientation, even under conditions of motion blur.
This development marks a significant step towards creating a reliable, low-light navigation system.
The research findings suggest that this orientation sensor could serve as a backup method for stabilising satellites and assisting drones and robots in navigating during night-time operations.
“Insects have been solving navigational problems for millions of years, including those that even the most advanced machines struggle with,” said Professor Chahl.
“And they’ve done it in a tiny little package. Their brains consist of tens of thousands of neurons compared to billions of neurons in humans, yet they still manage to find solutions from the natural world.”
A video explaining the research is available here, while the full study is accessible here.