This tiny chip sees like an eye and thinks like a brain

The team, led by Professor Sumeet Walia, states that the device can perform real-time and energy-efficient processing in many areas, from autonomous vehicles to robots.

Unlike conventional digital systems, this device uses an analog processing system that mimics the brain's working principle, allowing it to process complex visual data with much less energy.

“This technology operates at much lower power than digital systems used today and can perform complex visual tasks,” said Walia, director of the RMIT Centre for Optoelectronic Materials and Sensors.

At the heart of the device is molybdenum disulfide (MoS2), a material just a few atoms thick that can detect light and convert it into an electrical signal. It acts like neurons in the human brain, processing visual information in real time.

In the studies, the device was able to detect hand movements without capturing them frame by frame. This “edge detection” method works with less data and provides faster results. The device can store these movements in memory, just like a brain.

In addition to previous tests of the device with ultraviolet light, the research team also achieved successful results in the visible light spectrum.

This neuromorphic system could make vital decisions much faster, especially for autonomous vehicles and robots operating in dangerous environments. In the future, the device could pave the way for more natural robot interactions that can respond quickly to human movements.

The team is now working to expand the device from a single-pixel model to a larger pixel array, and is also developing versions of the device that can detect infrared light. This could enable real-time tracking of global emissions or toxic gases.

“Our aim is to develop hybrid systems that complement digital systems, are energy efficient and capable of real-time visual processing,” said Professor Walia.