Pose Estimation Strategy and Performance Analysis for Underwater Vehicles in Low-Visibility Environments

Underwater robotics is an emerging field of research with great challenges for achieving high levels of autonomy. By high levels of autonomy, one can consider the usage of submerged vehicles in unstructured environments to reach a georeferenced position while coping with obstacles and changes in the environment without any human intervention. A fundamental task in this scenario is determining the vehicle's accurate position and orientation so that the planning and perception capabilities can use this information for perfecting their output. This paper addresses the design and implementation an underwater vehicle localization architecture based on the Extended Kalman Filter (EKF) state estimation and multi-sensor data fusion capabilities. Two different use-cases are analyzed and a validation strategy for underwater vehicles, subject to low-visibility and high turbidity scenarios is. The performance obtained during sea trials is presented and an extension of the strategy to various submerged or surface vehicles with different sensor configurations is described. The novelty of the approach consists in the validation of the pose estimation algorithm on underwater vehicles and in the scalability of the proposed architecture to any type of submersibles.