A biorobotic approach to modelling celestial cue integration in ball-rolling dung beetles

This project aims to explore biologically plausible computational models which
may explain the mechanisms by which dung beetles can maintain a straight bearing
with reference to multiple sky cues as a compass reference. The initial approach
is centred on producing an accurate computer simulation of the dung beetle dorsal
eye view when presented with different patterns of light with the hypothesis that
the structure of the beetle eye and immediate optical pathway governs the cue
integration and cue preference differences observed between species. This
preprocessing will be used to generate a plausible encoding which can be used as
input to the Central Complex model; a model of the Central Complex neural
structure, thought to be the orientation centre of the beetle brain. While a
computational model is the goal, I aim to implement any resultant model on a
robot platform to provide performance data under real skylight conditions.

The question arises from studies in behavioural biology conducted at Lund
University regarding South African ball-rolling dung beetles; specifically, their
orientation behaviour when rolling dung away from the dung pile. Ball-rolling
dung beetles use dung for a few purposes, notably as a food source.
There is therefore serious competition over the valuable resource. Once a beetle
has formed its ball of dung it picks a direction and rolls the ball away in a
straight line to quickly maximise its distance from the dung pile.

Dung beetles (like humans) need some external reference to maintain their
bearing. Mounting evidence shows that the reference in use is the sky. Dung
beetles were first shown to use the sun as a directional prompt by Byrne et al.
in 2003 but have since been shown to use a variety of celestial cues to maintain
their straight-line orientation, including: the position of the Moon/Sun, a
celestial colour spectrum, the pattern of light polarisation caused by Moon/Sun
light filtering through the atmosphere, and even the the Milky Way galaxy in the
sky (visible from their natural habitat). Currently, we understand that the
available cues are combined in some way and that there is some inherent
preference for different cues between different species; however, there is no
biologically plausible computational model which adequately describes all of the
observed behaviours; this project aims to develop and explore such models.