State Estimation for Shore Monitoring Using an Autonomous Surface Vessel

Although many applications of small Autonomous Surface Vessels rely on two-dimensional state estimation, inspection tasks based on long-range sensors require more accurate attitude estimates. In the context of shoreline monitoring relying on a nodding laser scanner, we evaluate three different extended Kalman filter approaches with respect to an accurate ground truth in the range of millimeters. Our experimental setup allowed us to track the impact of sensors noise, including GPS non-Gaussian error, a phenomenon often underestimated. Extensive field experiments demonstrate that the use of a complementary filter in combination with a model-based extended Kalman filter performed best and reduced velocity errors by 73% compared to GPS. Finally, following our state estimation observations, we present a long-term shore monitoring result highlighting changes in the environment over a period of 6 months.