Neuromorphic Tactile Skin Sensors and Electronic Skin (E-Skin) Neural Networks: Flexible Piezoelectric Transduction, Spiking Tactile Processing, and Prosthetic Somatosensory Feedback

Advanced robotic manipulation, biomimetic prosthetics, and wearable health monitoring require highresolution electronic skin (e-skin) capable of simultaneously detecting static normal pressure, dynamic shear stress, surface texture micro-vibrations, and thermal gradients across large flexible surfaces. While traditional electronic skin sensor arrays utilize rigid silicon microcontrollers or continuous analog multiplexing, their deployment is severely constrained by massive wiring complexity, high static power consumption, slow frame-based sampling latencies, and mechanical fragility under repetitive stretching and bending. Neuromorphic Tactile Skin establishes a transformative bioinspired paradigm that integrates flexible piezoelectric and capacitive transduction arrays with asynchronous spiking neural networks (SNNs) implemented on flexible neuromorphic substrates. By emulating biological mechanoreceptors (SAI, SAII, RAi, PC units), e-skin sensors convert mechanical stimuli directly into sparse, event-driven spike trains with sub-millisecond temporal resolution and extreme energy efficiency. This paper delivers an exhaustive, mathematically formal technical investigation and empirical benchmark of neuromorphic electronic skin. We formalize electromechanical transduction via viscoelastic constitutive models, derive spiking mechanoreceptor spike-frequency adaptation equations, and construct multi-layer spiking neural networks for texture discrimination and slip detection. Furthermore, we construct an enterprise flexible electronics and hardware-in-the-loop testbed evaluating multi-modal e-skin patches across robotic tactile grasping, Braille character recognition, and closed-loop prosthetic somatosensory stimulation. Our empirical findings demonstrate that our neuromorphic e-skin achieves a 98.6% texture classification accuracy while reducing wiring density by 95.2% (via Address-Event Representation) and operating within an ultra-low power envelope of 0.84 mW per square centimeter. Finally, we formulate open challenges in large-area stretchable interconnect reliability, multimodal tactile-thermal sensory fusion, and direct neuro-prosthetic peripheral nerve interfacing.