Neuro-inspired tactile system for robotic embodied perception

Comprehension of real-world physical properties is a prerequisite for robotic embodied intelligence, crucially supported by advancing human-like tactile sensing. Though large-scale tactile sensing arrays are indispensable for improving resolution and accuracy, they are hindered by insurmountable scalability barriers—signal crosstalk and the overwhelming redundancy of data acquisition systems. In contrast, biological tactile perception systems can simultaneously and efficiently transmit and process stimuli from multiple receptors with distinct functions. To bridge this gap, we report a neuro-inspired tactile system integrating multiparametric sensing, neuromorphic encoding, and a cognitive framework to construct a complete sensation–perception–decision pathway. Specifically, we engineer a high-resilience aerogel/sponge composite structure featuring high sensitivity (7.54 kPa −1 ) and exceptional cyclic stability, along with self-decoupling thermal sensing (0.1 K resolution; 20.85 μV·K −1 sensitivity). This is paired with an ultrasensitive (21.86 kPa −1 ) toughened conductive thin film offering an ultrafast response (1.3 ms). By integrating these core sensors and spatial frequency encoding circuits into distinct spatial hierarchies to mimic the mechanism of the biological spinal tactile nerves, we realize signal-separable bus transmission and effectively emulate the typical firing dynamics of fast-adapting, slow-adapting, and thermal receptors. Finally, a large language model-driven iterative model is introduced, which synthesizes multimodal data to achieve nuanced comprehension of physical properties (compliance, roughness, and thermal insulation) and establish personalized classification criteria aligned with subjective user perception (94% accuracy). Real-world applications demonstrate that this holistic biomimetic design endows the robots with sensory acuity to extract rich multimodal information, laying a foundation for enhancing their understanding of physical properties.