On slam-based vehicle navigation with side-scan sonar data for an AUV

To ensure security for underwater installations or to protect certain underwater areas it is of utmost importance to obtain data about the structure of the environment. This should be done in a cost-efficient manner in an automatic way, e.g., by an autonomous underwater vehicle (AUV). Using only the on-board navigation sensors of an AUV for localization, noise is adding up until at a point the navigation data and the resulting maps are completely corrupted. Therefore, it is necessary to aid the vehicle navigation by additional measurements to constrain the navigational error during long-lasting missions. For surface vessels this is achieved by global navigational satellite systems (GNSS). As such systems are unavailable under water, alternative means to aid localization like, e.g., simultaneous localization and mapping (SLAM), need to be employed. In this article we describe how side-scan sonar sensors can be used as a device for sensing the environment in a SLAM context. It is important to treat the sonar measurements in such a way that a correct shape representation of the environment is obtained. Given the shape as an elevation map of the environment, salient parts of the environment have to be identified as it is essential for SLAM that those parts can reliably be found again upon a re-visit of the area. The Fraunhofer project TIETeK has created a prototype vehicle which applies these techniques to deep-sea environments.