Correcting Non-Uniform Sensitivity in EIT Tactile Sensing via Jacobian Vector Approximation

Electrical impedance tomography (EIT)-based tactile sensors enable promising capabilities for safe human-robot interaction through large-area distributed force sensing. However, their practical realization is hampered by non-uniform sensitivity distribution which varies at different locations. This prevents reliable force calibration and measurement. We propose a novel Jacobian vector correction (JVC) method to address this critical challenge. The JVC method corrects the spatial sensitivity of EIT sensors by constructing a tailored scaling vector derived from the system's Jacobian matrix. This achieves approximately uniform sensitivity across the entire sensing area without extensive calibration data requirements. We evaluated the JVC method through detailed FEM simulations which provided insights into the working principles. Phantom experiments using test objects with various conductivities validated its robustness. The method was further integrated into an EIT sensor prototype based on a porous polymer and ionic liquid structure. Tests showed accurate and consistent touch force prediction from 0 to 4 N across diverse locations on the circular tactile sensor with a diameter of 10 cm. The proposed JVC approach represents a significant advance in realizing the promising capabilities of EIT-based large-area tactile sensors across diverse human-robot interaction scenarios requiring sensitivity to safe touch forces.