SEOUL, May 12 (AJP) - Researchers at the Korea Advanced Institute of Science and Technology developed a sensor capable of adjusting its own response to light to overcome the limitations of current autonomous driving vision systems, the prominent institute said Tuesday.
Standard image sensors primarily detect light brightness, which can make it difficult for autonomous vehicles to distinguish between objects with similar reflectivity, such as water and asphalt at night. The new "self-reconfigurable" sensor array instead utilizes polarization, which is the direction in which light vibrates, to identify the surface structure and orientation of objects.
Korea Advanced Institute of Science and Technology (KAIST) Professor Seo Jun-ki and his team in the Department of Chemical and Biomolecular Engineering used a combination of tellurium and rhenium disulfide to create a specialized structure. This arrangement allows the sensor to change its operational state based on incoming light waves without requiring external electrical signals.
The team employed a precise manufacturing process called epitaxial atomic layer deposition to stack these materials at the atomic level. This method ensures the crystal structures align correctly, providing higher stability and more reliable performance compared to existing 2D material layering techniques.
This technology is designed for in-sensor computing, where the device processes visual data directly rather than sending it to a separate processor. In laboratory tests, the system achieved over 95 percent accuracy in recognizing moving objects while maintaining high energy efficiency.
"This research presents a new foundation for artificial intelligence vision technology that can secure richer visual information by using polarization," Professor Seo said. He noted that the technology is expected to play a significant role in developing low-power AI systems for autonomous driving and medical diagnostics.
Researchers Wenxuan Zhu and Kim Chang-hwan participated as lead authors of the study. The findings were published in the journal Nature Sensors on April 14, 2026.
(Reference Information)
Journal/Source: Nature Sensors
Title: Self-reconfigurable polarization perception in dual-anisotropy heterostructures for high-dimensional in-sensor computing
Link/DOI: https://bit.ly/4dkdYHE
Copyright ⓒ Aju Press All rights reserved.