A Brain‐Inspired ADC‐Free Vibrotactile System with Planar MoO ₃ Memristive Device Integrated Sensing‐Memory‐Computing Architecture

ABSTRACT Bio‐inspired vibrotactile perception is crucial for autonomous robotics, yet its development is fundamentally bottlenecked by the severe structural mismatch between traditional microscopic electronics and macroscopic flexible sensors, as well as signal complexity and the huge energy costs. Here, an integrated‐sensing‐memory‐computing (ISMC) architecture enabled by an interface‐oriented planar molybdenum trioxide (MoO 3 ) memristive device is reported to overcome these challenges. Featuring macroscopic lateral electrodes, this planar MoO 3 substantially alleviates the interface mismatch associated with conventional vertical memristors and enables direct circuit‐level electrical coupling with a self‐powered triboelectric nanogenerator (TENGs)‐based vibrotactile receptor through a full‐wave rectifier bridge. Driven intrinsically by the frequency‐dependent synaptic plasticity of the MoO 3 film, the system acts as a physical neural network that extracts features directly from raw analog vibration signals without analog‐to‐digital conversion (ADC). In practical industrial fault diagnosis, the system identifies five complex motor operating states with a 98.0% classification accuracy, matching software‐based Convolutional Neural Networks (CNNs). Remarkably, this hardware‐level processing achieves an ultra‐low average power of 0.250 µW—five orders of magnitude lower than typical digital hardware. This work demonstrates an architecture‐enabled pathway for coupling self‐powered mechanosensors with neuromorphic computing components, offering an energy‐constrained, ADC‐free paradigm for future embodied robotic perception.