Dr Pedro Machado, Nottingham Trent University

The challenge

AI systems can recognise images with high accuracy, but typically require substantial computing power, large datasets and high energy use. Biological vision is more efficient. The retina, the light-sensitive tissue at the back of the eye, detects motion, contrast, direction and sudden changes before sending information to the brain. Engineering systems that reproduce some of those early visual processes could help machines respond faster, operate with lower energy demand and work reliably at the edge, where data is collected. 

The research

Through the RAEng Leverhulme Trust Research Fellowship, Dr Machado will investigate neuromorphic computing, which uses artificial circuits and algorithms inspired by the nervous system, to emulate the retina. REACT will model five retinal motion-processing mechanisms: direction detection, object-motion sensitivity, looming detection, ON-OFF transient responses and motion anticipation. The project will use spiking neural networks, which communicate using short electrical-like pulses, and field-programmable gate arrays, which are reconfigurable hardware devices designed for efficient real-time processing.  

REACT aims to create and validate an energy-efficient prototype for motion perception, object tracking and collision prediction. Project outputs will include open-access models, datasets and a hardware demonstrator, supporting future research in robotics, intelligent sensing, assistive technology and longer-term bionic vision systems. 

The researcher

Dr Machado’s research focuses on neuromorphic computing, spiking neural networks, edge artificial intelligence and machine perception. His work is motivated by the efficiency of biological nervous systems and by the opportunity to translate those principles into practical technologies for safer, more adaptive and more sustainable intelligent machines.