Stretchable and energy‐efficient neuromorphic devices based on ionic capacitance for robotic tactile recognition

Abstract

Abstract Neuromorphic devices with learning and memory functions are critical for enabling human‐like tactile perception and interactive control in humanoid robots. However, existing neuromorphic platforms often struggle to reconcile mechanical flexibility with energy efficiency due to material and mechanism constraints. Here, we report an ion‐driven capacitive neuromorphic device with high stretchability, using a crosslinked polymer network ionic gel as the functional layer and liquid metal as electrodes. The device achieves ultra‐low energy consumption of 47 fJ per spike and excellent mechanical durability, maintaining stable postsynaptic response after 10,000 cycles of 100% strain stretching. The study further demonstrates that in a feed‐forward system, threshold‐triggered robotic grasping enables input‐history‐dependent response modulation and training‐state‐dependent tactile discrimination. This work provides a new design paradigm for low‐power, flexible neuromorphic systems, laying a solid foundation for potential applications in humanoid robotics and wearable electronics.