The CUSLAM project aims at the development of a vision-based localization and mapping algorithm (SLAM) specially tailored for underwater vehicles which operate in confined spatial structures. Traditional localization methods for underwater vehicles expect open terrain and constant line of sight to a grid of underwater localization transponders, which cannot be guaranteed when operating near or even in complex ground structures (i.e. underwater production facilities, riffs). A vehicle-centered vision-based approach can alleviate this problem. In order to validate the accuracy of the developed algorithm, a second localization measurement is made using a traditional setup, which then provides the gold-standard for comparison.
This presentation will give an overview on the current state of the project (running for 23 months now), including the custom build underwater vehicle "Dagon" and its instrumentation, the state of sensor fusion for the reference localization implementation, and current data from real vehicle tests in a lake environment.
