A/Prof Steven Wiederman- From neurobiology to autonomous robots: developing bio-inspired sensing and information processing system

Whether a human catching a ball, a dog leaping at a Frisbee or a dragonfly pursuing prey amidst a swarm, brains both large and small have evolved efficient solutions to tasks that challenge the most sophisticated robotic vision systems – the detection, selection and pursuit of moving targets in cluttered environments. With electrophysiological recordings from flying insects, we examine the brain mechanisms underlying target selectivity, predictive coding of trajectory, selective attention and roles for active-vision strategies. We develop computational models of this neuronal processing, simulating closed-loop pursuits in virtual reality environments. We also develop modified deep learning architectures, for applications in artificial vision systems. We translate these neuro-inspired models onto our autonomous vehicles, testing their ability to move and interact with features, within unstructured environments.