Abstract
Because of increasing demands for robots designed to work in the living environment and intuitive interfaces such as iPhones, it is necessary to develop high performance tactile sensors. We are developing a total-internal-reflection based tactile sensor, which reconstructs the sensor surface from a set of feature points on a reflection image of a image pattern. Our proposed method solves Geometric optical constraints by total optimization to reconstruct the sensor surface. By reconstruction from simulated images, we found the proposed method could perform high accurate reconstruction in real time and it was applicable for various sensor shapes and robust to tracking error. Then, we implemented the proposed method on the actual sensor and confirmed the real-time reconstruction.
