Learning-based configuration estimation of a multi-segment continuum robot

In this paper, we present a visual learning algorithm for estimating the configuration of a multisegment continuum robot designed for surgery. Our algorithm interpolates a stereo visual feature descriptor manifold using Radial Basis Functions (RBFs) to estimate configuration pose angles. Results are shown on a 3-segment snake robot, where rotational accuracy in the range of 1° -2° is achieved.