Machine Learning-Assisted Multi-Stage Highly Sensitive Electronic Skin Tactile Sensing Tracking Platform
The hierarchical spatial distribution of Merkel cells in the epidermis and Ruffini endings in the dermis provides a design paradigm for decoupling normal stress, shear stress, and strain in synthetic electronic skin. Existing biomimetic systems, however, struggle to replicate this three-dimensional mechanoreceptor topology while maintaining independent multimodal sensing. Zhang and co-workers report a 3D electronic skin (3DAE-Skin) that employs an eight-arm cage-like mesoscopic structure (height 600 μm) and an arched mesoscopic structure (height 250 μm) to spatially arrange force and strain sensing elements. A gradient modulus encapsulation strategy embeds force transducers in high-modulus polydimethylsiloxane (PDMS) and strain sensors in low-modulus Ecoflex, mimicking collagen fiber networks and dermal matrix mechanics, respectively. The resulting five-polyimide-dielectric-layer, two-force-sensing-layer, and two-strain-sensing-layer heterostack enables a 5×5 sensing unit array to achieve mechanical decoupling of normal force, shear force, and strain through piezoresistive transduction. The platform integrates machine learning algorithms to enhance multi-point tactile perception and maintain robustness under partial sensor failure. This work establishes a viable route toward high-fidelity tactile tracking for robotics, healthcare, and human-computer interaction, though challenges in temperature-humidity integration, self-healing, and algorithmic resilience remain. The reported architecture offers a concrete pathway for next-generation intelligent devices requiring spatially resolved, multimodal tactile feedback.