A Novel Particle Filter Formulation with Application to Terrain-Aided Navigation

Abstract

Terrain-aided navigation (TAN) holds high potential for integration with classical dead-reckoning methods in underwater vehicles that do not have access to the global positioning system (GPS). The TAN approach constitutes an economical solution for long-range oceanographic missions that cannot afford the cost of high-grade inertial navigation systems or the logistics associated to the deployment of long acoustic baselines. Successful implementations of the terrain navigation concept require efficient and robust algorithms for estimation of the vehicle kinematic states. Current implementations of terrain-based navigation rely essentially on a class of nonlinear filters commonly designated particle filters (PF) due to their versatility and robustness. However, there are some typical problems posed by TAN applications that require further investigation since they are not adequately solved by standard PF algorithms. The present paper addresses the problem of efficiency and robustness of particle filters in the context of terrain based navigation. To this effect, we present two new filter versions that are shown to be more robust and to achieve superior performance in terms of position and velocity estimation. When applied to TAN, the new PF formulations mitigate filter divergence problems frequently caused by terrain symmetries and are more robust than other well-known versions when used in scenarios with poor terrain information.