Synthetic Eyes Offer New Outlook for Robotics
No matter how much technology leapfrogs on a nearly fortnightly basis, autonomous vehicles equipped with advanced cameras and artificial intelligence remain surprisingly fragile when encountering high-contrast lighting environments. Sudden flashes of light, such as oncoming headlights piercing a dark night sky, can trigger data glitches that temporarily blind a vehicle’s system.
To solve this, a team of researchers from Pennsylvania State University (PSU), co-led by Larry Cheng, has developed a human-eye-inspired device that adapts to changing light conditions in seconds.
The breakthrough utilises a new type of “photomemristor,” a microscopic electrical component that simultaneously senses light and stores it as data, mimicking how neurons work.
The human eye relies on rod cells for low light and cone cells for bright conditions, using a biological “bleaching” process to manage sudden transitions.
To mimic this self-regulation, the PSU scientists paired titanium oxide with a stretchy, gel-like plastic known as PEDOT:PSS. The titanium oxide converts ambient light into an electrical current, which dictates how much water vapour the plastic absorbs from the air. The material rapidly swells in the dark and dries out in bright light, acting as a physical volume knob that automatically adjusts light sensitivity.
To evaluate the technology, the team integrated a miniature 4-by-4 grid of these photomemristors with a neural network. When challenged to distinguish a dimly lit letter “F” against a highly contrasting backdrop, the system achieved over 95% pattern recognition accuracy after just seven training cycles.
Cheng thinks this technology could lead the way to more sophisticated optical devices, providing adaptive and reliable sight back to patients with impaired vision. It could also allow for systems that result in robots operating in low-light conditions.
