Self-Powered Tactile Sensor with Learning and Memory

TLDR

Fabrication of human‑like intelligent tactile sensors is an intriguing challenge for developing human–machine interfaces. The study demonstrates intelligent neuromorphic tactile sensors with learning and memory based on a triboelectric nanogenerator, inspired by somatosensory signal generation and neuroplasticity. The sensors, built on a triboelectric nanogenerator, actively generate signals of varying amplitudes that depend on prior pressure stimuli, mimicking synaptic potentiation and memory, and are integrated into smart fingers that record both current and past actions. The sensors can retain memorized information for adjustable durations, allowing cascaded devices to exhibit multilevel forgetting and store abundant data, and they can serve as functional elements for artificial intelligence.

Abstract

Fabrication of human-like intelligent tactile sensors is an intriguing challenge for developing human–machine interfaces. As inspired by somatosensory signal generation and neuroplasticity-based signal processing, intelligent neuromorphic tactile sensors with learning and memory based on the principle of a triboelectric nanogenerator are demonstrated. The tactile sensors can actively produce signals with various amplitudes on the basis of the history of pressure stimulations because of their capacity to mimic neuromorphic functions of synaptic potentiation and memory. The time over which these tactile sensors can retain the memorized information is alterable, enabling cascaded devices to have a multilevel forgetting process and to memorize a rich amount of information. Furthermore, smart fingers by using the tactile sensors are constructed to record a rich amount of information related to the fingers' current actions and previous actions. This intelligent active tactile sensor can be used as a functional element for artificial intelligence.