Recently, a team led by Zhou Feichi from the School of Microelectronics at the Southern University of Science and Technology (SUSTech) has made remarkable strides in the realm of wide field-of-view monocular 3D perception computing. Their significant findings have been published in Nature Communications. This groundbreaking study introduces, for the first time, a wide field-of-view anisotropic optically controlled resistive random-access memory (AORRAM). This innovation enables in-situ wide field-of-view motion perception and depth estimation directly within the sensor. Additionally, the team has constructed a compact, low-power system for wide field-of-view monocular 3D motion perception and localization.

Drawing inspiration from the wide field-of-view, efficient 3D perception, and motion tracking abilities of eagle eyes, the researchers designed and fabricated a wide field-of-view (140°) AORRAM. This device is based on zinc oxide nanowire arrays featuring an axial oxygen vacancy concentration gradient. By integrating this device, they have successfully built a 3D perception and motion localization system on a unified hardware platform, effectively mimicking the exceptional visual sensing and computing capabilities of eagle eyes.

Experimental results reveal that the system can accurately determine the distance and azimuth of moving targets in real time. This allows for precise acquisition of depth and positional information, achieving an impressive 3D motion recognition accuracy of 95.41% in complex scenarios. The AORRAM device and system proposed in this study offer substantial advantages over traditional 3D machine vision systems, particularly in terms of power consumption, latency, and system compactness. These advancements hold great promise for driving the application of next-generation compact, high-energy-efficiency 3D intelligent vision systems in various fields, including autonomous navigation of unmanned aerial vehicles, agile robotics, and real-time environmental interaction.