Toward robust localization and mapping for shallow water ROV inspections

We propose a portfolio of filtering methods for a remotely-operated vehicle (ROV) that relies on a compass, gyro, depth sensor and ultra-short baseline (USBL) positioning system for localization in shallow-water environments. Our goal is to maintain an accurate state estimate that will be suitable as a basis for decision-making in the course of carrying out an autonomous underwater inspection. We employ Kalman filters in conjunction with an outlier filter to produce a reliable estimate in the presence of noisy and spurious data. Our localization result is then used as a basis for 3D occupancy mapping, in which further filtering and inference techniques are applied to remove false returns from the ROV's scanning sonar. Localization results from two field deployments of the ROV are given.