A reduced-order-model Bayesian obstacle detection algorithm

We consider an efficient Bayesian algorithm for solving the inverse problem of locating and classifying certain obstacles using noisy far field data obtained by illuminating them with a wave. The efficiency of our approach comes from using a reduced order model for the wave propagation problem based on the T-matrix. The key to assembling the reduced order model is a novel way to compute the T-matrix using far field data, compared with the standard approach using near field. This novel approach greatly extends the kinds of obstacle that can be simulated and detected.