Tactile Near‐Sensor Analogue Computing for Ultrafast Responsive Artificial Skin

Ultrafast artificial skin enables unprecedented tactile internet applications in prosthetics, robotics, and human-machine interactions. However, current artificial skin systems that rely on front-end interface electronics typically perform redundant data transfer and analogue-to-digital conversions for decision-making, causing long latency (milliseconds). Here, a near-sensor analogue computing system based on a flexible memristor array for artificial skin applications is reported. This system, which seamlessly integrates a tactile sensor array with a flexible hafnium oxide memristor array, can simultaneously sense and compute raw multiple analogue pressure signals without interface electronics. As a proof-of-concept, the system is used for real-time noise reduction and edge detection of tactile stimuli. One sensing-computing operation of this system takes about 400 ns and consumes on average 1000 times less power than a conventional interface electronic system. The results demonstrate that near-sensor analogue computing offers an ultrafast and energy-efficient route to large-scale artificial skin systems.