A strain-sensitive neuromorphic device emulating mechanoreception for different skin sensitivities

Abstract

Emulating the somatosensory cognitive ability of the human body with neuromorphic devices is an upcoming activity. Among multiple domains, tactile sensors or mechanoreceptors have particularly captivated a lot of interest due to their potential to detect and measure physical interaction. While there have been reports on using strain sensors coupled with neuromorphic devices to perform such actions, sensors with a built-in ability to sense are yet to be demonstrated. Here, we report a study on the fabrication of a neuromorphic device that makes use of the inherent strain-sensing mechanism coupled with neuromorphic functions. This was accomplished using an interconnected network of gold microwires embedded in polydimethylsiloxane (PDMS), which exhibited sensitivity to the applied strain. The device demonstrated an inherent change in resistance when mechanical strain was applied. The relaxation after strain removal was carefully monitored and found to follow the Ebbinghaus forgetting curve. Various neuromorphic functionalities like short-term plasticity (STP), long-term plasticity (LTP), spike rate-dependent plasticity (SRDP), spike amplitude-dependent plasticity (SADP), potentiation, and depression have been demonstrated. The device showcased remarkable performance with high linearity (non-linearity factor as 0.29 for potentiation and −0.09 for depression) and paired-pulse facilitation (PPF) levels (232%) approaching those found in biological systems. Furthermore, by strategically modulating the Young's modulus of the PDMS, the mechanoreception was extended to real skin-like conditions with varying sensitivities, such as that found in tongue and lip areas in contrast to the hard sole. Our observations on the impact of this modulation on device performance provide unprecedented insights, marking a pioneering advancement in artificial sensory systems.