Moth Flight Patterns Help Drones Navigate Complex Environments

Researchers from Boston University (BU) have captured data from moth flight patterns to improve drone navigation through complex environments.

To understand how moths plan their route, the researchers, led by Ioannis Paschadilidis at BU and Thomas Daniel at the University of Washington, mounted eight hawk moths (Mantuca sexta) on metal rods connected to a torque meter. In front of each moth they projected a moving forest scene created from beams of light for the moths to navigate.

An image of the experimental setup showing a moth attached to a metal rod in front of the virtual forest scene. Courtesy of Thomas Daniel lab, University of Washington.

The researchers found that the moths mainly rely on the pattern created by the apparent motion of objects caused by their flight, which agrees with studies of flight behavior in other insects. The researchers then used the data to create a mathematical model to describe the moth’s trajectory through the forest. The data was then translated into a decision-making program that could be used to control a drone. Then the researchers compared how the drone and the moth performed in simulations of the same forest layout, as well as new configurations with different densities of trees.

The drone patterns performed 60% better in the simulated forests because they also incorporate information about the exact location of objects in their surroundings into their navigational decisions.

Though the researchers were able to improve performance in certain environments, the moths’ strategy was more adaptable, performing well in a variety of different forest layouts. The moth model performed best in dense forests, suggesting that hawk moths have evolved a flight strategy adapted to the thick forests they encounter.

The researchers say that by using further data from animal flight paths, they can program bio-inspired drones that will be able to navigate autonomously in cluttered environments.

The research was published in PLOS Computational Biology (www.doi.org/10.1371/journal.pcbi.1007452).