Underwater Geophysical Navigation using a Particle Filter Approach to Multi-Sensor Fusion

This paper addresses the problem of underwater navigation of autonomous underwater vehicles via a Monte Carlo estimation approach that relies on the use of a prior digital elevation map of the seabed and bathymteric data acquired with a Multibeam Echosounder. The Monte Carlo estimation procedure is implemented in the form of a particle filter, as part of a multi-sensor fusion framework for underwater geophysical navigation. The contribution of this paper focuses on two main topics: i) development of a particle filter as a solution to a terrain-based Bayesian vehicle positioning problem, followed by filter implementation and simulation and ii) integration of the particle filter structure in a multi-sensor fusion architecture for vehicle navigation with a view to increasing navigation accuracy. The results of realistic simulations with a dedicated marine robotics system simulator illustrate the navigation performance achieved with the proposed solution.