The pursuit of endowing robots with human-like perception and intelligence has reached a meaningful milestone.Researchers at an Australian university have engineered a neuromorphic device that mirrors the human brain’s real-time visual information processing. This breakthrough holds the potential to revolutionize robotics, artificial vision, and autonomous systems.

Brain-Inspired Technology

The human brain’s visual processing relies on a complex network of neurons communicating via electrical signals, frequently enough referred to as “peaks” or impulses. to emulate this mechanism, scientists have developed spiking neural networks (snns) that simulate neural behavior. A prominent model is the leaky integrate-and-fire (LIF) model.

In the LIF model, electrical signals accumulate in a neuron until a threshold is reached, triggering an impulse. The system then resets, ready for new information. Replicating this process on a microscopic scale within an electronic device and applying it to real-time vision presents a considerable technological challenge.

molybdenum Disulfide: The Innovation’s Core

The Australian team achieved this feat using molybdenum disulfide (MoS₂), a material with unique properties. MoS₂ is a metallic compound that, at the atomic level, possesses natural defects enabling it to detect light and convert it into electrical signals, mimicking the behavior of brain neurons that capture and transmit visual signals.

Using ultra-thin layers of mos₂ obtained through chemical vapor deposition, the researchers designed a device capable of simulating the electrical load and discharge of a neuron according to the LIF model. This system detects light variations,”turns on” upon reaching a threshold,and resets quickly through precise voltage control,functioning as a real neuron with rapid reaction capabilities.

Real-Time Vision and Memory

The device can detect movements, such as that of a hand, and store this information as temporary “memories.” Unlike conventional cameras that capture images frame by frame,this device directly analyzes contours and environmental changes,significantly reducing the data to be processed.

By reproducing the way the brain filters and analyzes visual information,this neuromorphic device thus consumes much less energy while providing real -time treatment.

University of Australia Research Team

This ability to react immediately to environmental changes opens new avenues in autonomous robotics and smart vehicles.

Towards More Reactive, Human-Like Robotics

This technology is particularly promising for humanoid robots that must interact naturally with dynamic and complex environments. A vision system closer to human vision, coupled with the ability to store and analyze visual information in real-time, would enable faster and more appropriate reactions, especially in rapidly evolving contexts.

A vision closer to that of humans, coupled with an ability to store and analyze visual information in real time, would allow faster and more suitable reactions, especially in rapid or rapidly evolving contexts.

University of Australia Research Team

Beyond robotics, this device could revolutionize autonomous driving, where rapid detection and analysis of visual information are crucial for safety. it could also be integrated into assistance systems for the elderly or disabled, making interactions more intuitive and effective.

future Perspectives and Challenges

The current prototype operates at the scale of a single pixel. The team is now working to create larger matrices of neuromorphic pixels based on MoS₂, capable of managing more complex images. Efforts are also underway to optimize energy consumption and integrate this technology into conventional digital architectures.

Researchers are also exploring other materials to extend detection capabilities to the infrared spectrum, potentially enabling applications in environmental surveillance or industrial emissions monitoring.