Terrain Modeling for Autonomous Underwater Navigation

Abstract

The main task of perception for autonomous vehicles is to build a representation of the observed environment in order to carry out a mission. In particular, terrain modeling, that is modeling the geometry of the environment observed by the vehicle's semors, is crucial for autonomous underwater exploration. The purpose of this work is to analyze the components of the terrain modeling task, to investigate the algorithms and representations for this task, and to evaluate them in the context of real applications. Terrain representation is an issue that is of interest in many areas of mobile robotics, such as land vehicles, planetary explorers, etc. This paper surveys some of the ideas developed in those areas and their relevance to the underwater navigation problem. Terrain modeling is divided into three parts: structuring sensor data, extracting features, and merging and updating terrain models.