In my PhD I aim to elucidate the algorithmic basis of pattern recognition in insects - particularly the hummingbird hawkmoth. I use computer vision tools to track their behaviour and movement while foraging, trying to understand (and predict) the neural processing employed when investigating pattern features.
Historically, animal behaviour research (and neuroscience generally) has largely drawn conclusions from simplistic stimuli presented to animals in unnatural, movement-restricted conditions. While useful to investigate the basic fundamentals of neural processing and simple behavioural sequences, the richness and complexity of organisms in these conditions can be lost. Many animals (alongside humans and even robots) actively explore their environments, seeking out a diverse array of stimuli across modalities which they use to make decisions. I find our current era of neuroethology extremely exciting because of the collective effort to combine more naturalistic stimuli with less invasive and restricting recording methods to further understand the brain and behaviour.
Drawing inspiration from an integrative biology approach, I aim to understand how organisms operate on a variety of scales - from the function of individual sensory receptors, to the physical interaction of the body with the environment. To combine these aspects in my PhD, I investigate pattern recognition in the hummingbird hawkmoth (Macroglossum stellatarum). I employ computer vision techniques to track moths non-invasively while they freely explore foraging sites in controlled environments. In nature, they use the patterns on flower petals to guide their hovering and nectar sipping. In the lab, I collect behavioural and kinematic information about how they actively use pattern features to guide such behaviours. I intend to use this information to understand the algorithms underlying their intriguing sensorimotor processing, and pattern vision overall. In doing so, we also aim to integrate ourselves with bioroboticists who are interested in biologically-inspired robots and neural networks for computer vision.