Heterogeneous Electronic Skin Integrated With a Neural Network for Simultaneous Sensing of Strain Direction and Intensity

Mechanically stretchable strain‐sensing electronic skin (E‐skin) has garnered significant attention for bioinspired sensory systems and human‐machine interfaces (HMI). However, conventional E‐skin devices or arrays remain limited in achieving high‐precision simultaneous detection of strain direction and intensity, which holds promise for enabling robots to recognize touch gestures, and even interpret intentions and emotions. To address this challenge, this study developed a heterogeneous E‐skin that integrates three stretchable strain‐sensing units with a backpropagation (BP) neural network algorithm. Each sensing unit features a heterogeneous fence‐structured design, which minimizes transverse strain interference and endows orientation sensitivity‐responding exclusively to longitudinal strains. The sensory layer exhibits excellent flexibility, stretchability, and cyclic stability, with a gauge factor (GF) of 2.96 under 0–50% strain, and 10.93 under 50–80% strain. Through combined simulation analysis and experimental validation, the heterogeneous structure was optimized to achieve superior directional strain response characteristics. Notably, by employing the BP neural network to decode correlated signals from the three sensing units, the E‐skin achieves high‐precision detection of both the direction and intensity of an arbitrary applied strain.