Diffusion-based outlier rejection for underwater navigation

This paper addresses the issue of rejecting spurious acoustic position measurements for the estimation of trajectories in underwater vehicle navigation. The method relies on the diffusion-based observer approach (Jouffroy and Opderbecke, 2004; 2005), which allows one to consider and process entire trajectory segments at a time. The proposed observer contains a nonlinear feedback gain, constraining the estimate to detach from measurements that are considered as outliers. The observer is proven to be exponentially convergent. Simulation and experimental results are presented to illustrate the benefits of the approach.