The hawk moth's wings are a blur of mottled grey motion as it hovers tethered to a steel rod in large white plastic orb. Outside the orb in the darkened room where I stand, a projector casts moving patterns of dimmed light onto the sphere's surface, illuminating the moth's field of vision with oscillating stripes.
Tonya Muller, a DPhil student in Oxford University's Department of Zoology, sits at the computer controlling the experiment. At regular intervals, she directs the computer to alter the direction, amplitude and frequency of the light stripes.
These changing light patterns create altered visual environments for the moth inside, which aim to simulate real-world visual disruptions the moth might experience when exposed to wind gusts. As the patterns change, the moth makes rapid adjustments to its flight behaviour to maintain constant stability.
Though imperceptible to the human eye, the moth's responses to the visual stimuli are detected by a force sensor attached to the end of the steel rod and relayed to Tonya's computer. These recordings are helping Tonya to understand the moth's remarkable visual-motor system, and identify the mechanisms of visual feedback in insect flight control.
'Understanding vision-based flight control in insects has far reaching uses in the fields of sensor development, signal processing, and robotics,' says Tonya, whose background is in mechanical engineering. Vision is important for information gathering in insects and up to 50% of an insect's brain can be composed of visual neurons. In fact, despite their small brain size, insects can solve extremely sophisticated orientation problems both rapidly and reliably. Yet their eyes are far less sophisticated than our own.
'Insects receive visual information through a relatively noisy, low-resolution sensor. But with this sensor they are able to processes information at sufficient speeds to react and respond to unexpected disturbances,' Tonya tells me.'This is extremely interesting from an engineering perspective because developing technologies that use simpler and fewer electrical sensors and perform equally well can reduce manufacturing costs and computational power.'