Electronic skin (e-skin) is emerging as a key enabling technology for humanoid robots, wearable robots, prosthetic systems, and embodied Physical AI. Unlike conventional tactile sensors designed primarily for localized pressure measurement, future robotic skins must simultaneously accommodate large mechanical deformation, distributed tactile sensing, multimodal perception, large-area electrical routing, and intelligent interpretation of physical interactions. However, no single material or structural architecture can optimally satisfy all of these requirements. A systematic framework comparing alternative e-skin architectures is therefore necessary for selecting appropriate sensing and interconnection strategies for different robotic applications.

This paper presents a comparative framework for four representative multimodal electronic-skin architectures intended for Physical AI systems: (1) latex–silicone-based auxetic stretchable printed-circuit e-skin, (2) CNT/graphene-based auxetic conductive e-skin, (3) hydrogel–liquid-metal–auxetic large-area electronic skin, and (4) an AI-integrated multimodal whole-body e-skin architecture. The first architecture employs geometrically deformable auxetic and serpentine conductive pathways combined with neutral-plane structural engineering to accommodate stretching, bending, twisting, pressure, and shear deformation. The second architecture utilizes carbon nanotube and graphene conductive networks to provide mechanically compliant and distributed strain-sensitive pathways. The third combines hydrogel interfaces, liquid-metal conductors, and auxetic geometries to achieve extreme deformability and large-area conformability. The fourth extends these material and structural technologies toward distributed multimodal perception by integrating tactile, proprioceptive, thermal, vibration, and deformation information with hierarchical artificial intelligence.

A unified electromechanical formulation is introduced to describe strain transfer, resistance variation, multimodal sensor response, sensor fusion, and AI-based tactile interpretation across the four architectures. Their relative characteristics are compared in terms of stretchability, bending and torsional compliance, pressure/shear sensitivity, electrical stability, self-healing potential, large-area manufacturability, signal-routing complexity, and suitability for Physical AI. Based on this comparison, a hierarchical design methodology is proposed in which different e-skin technologies can be selectively deployed over fingertips, hands, joints, limbs, and torso regions according to local mechanical and perceptual requirements.

Rather than identifying a single universally optimal electronic-skin technology, this work establishes the materials structure–sensing–intelligence design framework for future robotic skin. The proposed perspective provides a reference architecture for researchers developing whole-body tactile interfaces, humanoid robots, wearable robotics, intelligent prostheses, and embodied Physical AI systems.