RMIT engineers create a brain-inspired device that sees and thinks in real time, advancing robotics and autonomous tech.
Researchers have developed an advanced neuromorphic device mimicking how the human brain processes information, marking a major step forward in autonomous technology. The compact system, developed by a team at RMIT University in Australia, can detect hand movements, store memories, and process visual data in real time, without relying on an external computer.
This innovation paves the way for advanced robotics, including humanoids, autonomous vehicles, and next-generation systems designed for seamless human interaction.“This proof-of-concept device mimics the human eye’s ability to capture light and the brain’s ability to process that visual information, enabling it to sense a change in the environment instantly and make memories without the need for using huge amounts of data and energy,” said Sumeet Walia, professor and team lead, at RMIT Unviersity, in a statement. Brain-like sensingNeuromorphic vision and information processing are rapidly growing fields that aim to create smarter, more efficient computing and sensing systems. One key approach uses spiking neural networks , which work like real brain cells by triggering signals, or “spikes,” when triggered. A common model for this is called the leaky integrate-and-fire model. In this model, electrical signals build up until they reach a certain level, then a spike is sent out, and the system resets, just like real neurons behave. Although many light-sensitive materials have been tested for basic brain-like functions, accurately copying the full LIF behavior, especially how the system stores and resets its electrical state, and using it in visual tasks, is still a new and largely unexplored area.The device detects hand movement, stores memories, and processes information like a human brain, without needing an external computer.RMIT researchers combined neuromorphic materials with advanced signal processing to create a device capable of capturing and processing visual information in real time. At the core of the technology is molybdenum disulfide , a metal compound with atomic-scale defects that can be used to detect light and convert it into electrical signals, much like neurons in the human brain.The new research shows that ultra-thin layers of MoS₂, made using chemical vapor deposition, can mimic how brain cells charge and discharge, just like in the leaky integrate-and-fire neuron model. These layers respond to light in a way that lets them copy the electrical behavior of real neurons. Adjusting the gate voltage allows the system to quickly reset itself, helping it respond faster, just like a real brain. Smart visual sensingThe researchers built a spiking neural network using the key light-response features of MoS₂. This model achieved 75 percent accuracy on static image tasks after 15 training cycles and 80 percent accuracy on dynamic tasks after 60 cycles, showing strong potential for real-time vision processing.In experiments, the device detected hand movements using edge detection, avoiding frame-by-frame capture and reducing data and power use. It then stored these changes as memories, mimicking brain function. This work, done in the visible light range, builds on earlier research in the ultraviolet spectrum.“We demonstrated that atomically thin molybdenum disulfide can accurately replicate the leaky integrate-and-fire neuron behaviour, a fundamental building block of spiking neural networks,” said Thiha Aung, a PhD scholar at RMIT, and a first author of the study, in a statement.According to the team, previous UV-based work focused on still image detection, memory, and processing. UV and visible light devices can reset memories to prepare for new tasks.The innovation could significantly improve how autonomous vehicles and advanced robots respond to visual input, especially in high-risk or fast-changing environments. By detecting scene changes instantly with minimal data processing, the technology enables faster, more efficient reactions. This could also enhance human-robot interaction in areas like manufacturing or personal assistance. Researchers are now scaling the single-pixel prototype into a larger MoS₂-based pixel array, supported by new research funding. Plans include optimizing the device for more complex vision tasks, improving energy efficiency, and integrating it with conventional digital systems.The team is also exploring other materials to expand capabilities into the infrared range for applications like emission tracking and smart environmental sensing.The details of the team’s research were published in the journal Advanced Materials Technologies.
Nueromorphic RMIT University Robotics
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Tiny humanoid robot screw could hurt US and allies in hot bot race against ChinaChina strategically leads the supply of planetary roller screws, which is a critical component for advanced movement in humanoid robots.
Read more »
This Tiny Robot Is Part Bee, Part Crane Fly—and It Finally Has Solid LegsHarvard's RoboBee will one day conduct artificial pollination and survey disaster zones, but first it has to stop crash landing.
Read more »
Jeremy Renner's 'tiny but monumental' mistake that led to near-fatal accidentJeremy Renner revealed the 'critical' missed step that led to the near-fatal snowplow accident. The 'Avengers' star was run over by his snowplow on Jan. 1, 2023.
Read more »
Tiny Puppy Born on the Streets Manages to Save Mom and 10 Siblings'Sometimes destiny comes walking through a door,' commented one user.
Read more »
This Tiny Train Stuffed With Miniature Polos Might Be VW’s Coolest Birthday Tribute YetThe detailed scale model is exhibited at the Miniatur Wunderland in Hamburg, Germany
Read more »
World’s thinnest robot tram climbs wires, moves like screw to carry tiny payloadsAn infrared beam turns a twisted elastomer loop into a self-rolling cable-car soft robot that can carry 12 times its own weight.
Read more »