Biological-Fidelity, Knitted Computational Skin for Physical AI

Vision

We are looking to achieve biological-fidelity tactile and pressure sensing that mimics the complex mechanoreceptors in human skin while remaining computationally scalable and highly manufacturable. As AI expands to computational ‘brain’ power, there is a critical lack of physical sensing and perception capabilities that prevents robots from achieving a full loop of sense-compute-actuate and thus fully taking advantage of rich environmental data to learn and interact.

Human skin ranges from 50-200 mechanoreceptors per cm², a density of tactile and pressure sensors that would be astounding for physical AI systems to interact with and learn from their environment. Using knitting as an additive manufacturing technique that enables stitch-level tuning of fabric mechanical properties, we envision creating flexible, application-specific, and highly dense tactile sensors from our custom-made functional yarns composed of a conductive yarn core dip-coated with an insulation layer. Our dip-coating fabrication of insulated layers allows for highly dense looping of yarns and enables pressure-based capacitive sensing and Triboelectric Nanogenerator (TENG) sensing while preventing electrical shorts.

Leveraging our lab’s strength in computational engineering, our long-term vision is to build unique sensing architectures that facilitate simultaneous TENG and capacitive sensing in the same fabric layers and mimic both the fast-adapting (TENG) and slow-adapting (capacitive) mechanoreceptors in our skin. Unique architectures also deliver scalability in sensor computation and energy efficiency via hierarchical sensing. TENG layers act as wake-up signals that indicate a touch and relative location, prompting the capacitive layer to do a deeper scan of the touched area for precise force mapping and subtle tactile changes.

We aim to create manufacturable and scalable knitted sensing grids for human health applications such as prosthetic fit and force monitoring, biometric sensing, and athletic recovery/performance. We also see applications in robotic tactile skins for hands and body-scale sensing in manufacturing, caregiving, and dexterous manipulation applications.