An international group of scientists has developed a miniaturized photopamistrator that mimics the human eye’s ability to adapt automatically to sharp brightness swings. Research published in the journal Nature Communications suggests a solution to one of robotics' main problems: conventional cameras often "blind" under contrast lighting, when a drone or robot needs to simultaneously distinguish objects in dark zones and bright light sources, such as the headlights of oncoming vehicles. The new sensor allows machines to perceive the environment clearly in any conditions without loss of accuracy.

Device about 0.5 mm in size refers to neuromorphic systems, where the sensor not only captures an image but also partially processes a signal on site, which offloads the processor and accelerates the robot’s response. The secret of adaptation lies in the use of titanium dioxide and the PEDOT:PSS polymer, which react to environmental humidity. In low light conditions, the structure absorbs moisture, increasing its conductivity and photosensitivity, and in bright light, moisture evaporates, causing the sensor to reduce sensitivity so as not to "go blind."

During the experiment, researchers assembled a matrix of 16 such elements and connected it to an artificial neuronetwork that successfully recognized the letter patterns in mixed lighting conditions with an accuracy of 91.3%. This innovative approach is a great complement to the development of event cameras and opens new avenues for energy-efficient machine vision systems for autonomous transport and smart devices. It is worth noting that this is not the only recent breakthrough in AI: earlier this year, the scientific community also unveiled an autonomous Qumus system capable of conducting experiments with quantum materials on its own.