Penn State researchers have built a human-eye-inspired device to prevent self-driving cars from going blind in changing light conditions. 

It is a new type of ‘photomemristor’ that mimics the human eye, adapting to mixed, changing light from bright to dark in seconds.

“Self-driving cars are exposed to a mixture of light levels in use – imagine the contrast of the dark sky with the bright headlights of other cars when driving at night,” said Larry Cheng, James L Henderson Jr Memorial Associate Professor of Engineering Science and Mechanics at Penn State. 

“It can be difficult for an artificial optical system to distinguish details, like the glow of a red light, in these mixed lightning conditions,” the co-corresponding author added. 

Neural network experimental setup. Image: Jia Zhu.

Mimicking the human eye

Autonomous vehicles remain surprisingly fragile when encountering high-contrast lighting environments, despite being equipped with advanced cameras and powerful AI. 

These robotic eyes perform well in stable weather conditions, but are not accustomed to sudden flashes of light, such as oncoming headlights piercing a dark midnight sky. It can trigger catastrophic data glitches that blind the vehicle‘s system.

The team of engineers has finally found a workaround. And the secret is copying the wet, organic mechanics of the human eye.

Researchers unveiled a tiny hardware component that can adapt from blinding brightness to deep shadow in just seconds.

While our eyes require roughly 20 to 30 minutes to fully acclimate to extreme changes in light, this artificial system handles the transition almost instantly.

To build it, scientists had to rethink how computers process light. Standard cameras capture an image and send it to a separate computer brain for analysis. This takes time and immense computing power. The team used photomemristors. These are microscopic electrical components that can simultaneously sense light and store it as data – mimicking how neurons work.

But the key development lies in how the devices manage sensitivity.

4×4 grid for testing

The human eye uses rod cells for low light and cone cells for bright conditions. When a bright light hits your eyes, the pigments in your rods temporarily “bleach” out, leaving the cones to take over. To mimic this, the researchers paired a powdery white compound called titanium oxide with a stretchy, gel-like plastic known as PEDOT:PSS.

The titanium oxide captures ambient light and converts it into an electric current. That current forces the plastic layer to react with the surrounding air. In the dark, the plastic rapidly sucks in water vapor and swells. In bright light, it dries out quickly.

This automatic sweating and swelling acts as a physical volume knob for light. It self-regulates.

To test the tech, researchers built a miniature 4×4 grid and paired it with an AI neural network. They gave it a classic ophthalmologist challenge: spot a dimly lit letter 'F' against a massively bright backdrop.

The system didn’t flinch. It took just seven rounds of training for the device to reach a remarkable 95% accuracy in pattern recognition.

The implications stretch far beyond avoiding highway accidents. The team has already filed a provisional patent to expand their use.

In the future, factory robots could operate flawlessly in erratic, flickering industrial environments. More importantly, Cheng believes this technology could eventually yield sophisticated artificial optics, giving adaptive, reliable sight back to the visually impaired.

The findings were published in the journal Nature Communications on June 9.